CN111126153B - Security monitoring method, system, server and storage medium based on deep learning - Google Patents

Abstract

The embodiment of the invention discloses a safety monitoring method, a system, a server and a storage medium based on deep learning. The method comprises the following steps: acquiring video data of a target to be monitored in a monitoring area; analyzing the video data through a first model trained in advance to acquire monitoring characteristic information of the target to be monitored in a monitoring area; judging whether the target to be monitored is abnormal or not according to the monitoring characteristic information; and when the target to be monitored is abnormal, alarm operation is carried out. According to the invention, through training based on deep learning on video data, semantic structured extraction of the video data and effective management of the video data are realized.

Description

Security monitoring method, system, server and storage medium based on deep learning

technical field

Embodiments of the present invention relate to video image analysis technology, and in particular to a safety monitoring method, system, server and storage medium based on deep learning.

Background technique

At present, the video surveillance system has become an important means of application in related fields. Through video surveillance, monitoring personnel or security personnel can conduct security protection and investigation on the monitoring area more directly and effectively. However, due to the huge amount of video data and complex format, the storage is expensive and difficult to manage. Faced with massive video information, unstructured data forms and ambiguity in content, manual retrieval is time-consuming and labor-intensive. A large number of videos are lost without sorting out, which seriously affects the effectiveness of the monitoring system.

Contents of the invention

The present invention provides a safety monitoring method, system, server and storage medium based on deep learning, so as to realize semantic structured extraction of video data and effective management of video data.

In the first aspect, the embodiment of the present invention provides a safety monitoring method based on deep learning, including:

Obtain the video data of the target to be monitored in the monitoring area;

Analyzing the video data by using the pre-trained first model to obtain monitoring feature information of the target to be monitored in the monitoring area;

Judging whether the target to be monitored is abnormal according to the monitoring characteristic information;

When there is an abnormality in the target to be monitored, an alarm operation is performed.

Further, the monitoring feature information includes one or more of monitoring target action information, monitoring target staying time information, smoke detection information, and monitoring target clothing information.

Further, the video data is analyzed through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area, including:

The convolutional neural network is trained on the video data to obtain a first model corresponding to analyzing and monitoring target action information, monitoring target stay time information, smoke detection information and monitoring target clothing information.

Further, analyzing the video data through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area includes:

Using the first model to identify the video data based on the monitoring target action information, so as to obtain the monitoring target action information of the target to be monitored;

calculating a first confidence level of the monitoring target action information according to a first preset weight value;

The action state of the target to be monitored is judged according to the preset parameter threshold and the first confidence level of the action information of the monitored target.

Further, analyzing the video data through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area also includes:

Using the first model to identify the video data based on the duration information of the monitoring target, so as to obtain the duration information of the monitoring target of the target to be monitored;

Calculating the second confidence degree of the monitoring target staying time information according to the second preset weight value and the monitoring target staying time information;

The stay state of the target to be monitored is judged according to the preset parameter threshold and the second confidence level of the stay duration information of the monitored target.

Further, analyzing the video data through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area also includes:

Using the first model to identify the video data based on the smoke detection information, so as to obtain the smoke detection information of the target to be monitored;

Calculate the smoke concentration according to the smoke detection information and determine the smoking action of the area where the smoke is located and the target to be monitored;

Calculate the third confidence level of the smoke detection information according to the smoke concentration, the area where the smoke is located and the smoking action of the target to be monitored;

The smoking state of the target to be monitored is judged according to the preset parameter threshold and the third confidence level of the smoke detection information.

Further, analyzing the video data through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area also includes:

Using the first model to identify the video data based on the monitoring target clothing information, so as to obtain the monitoring target clothing information of the target to be monitored;

calculating the fourth confidence degree of the monitoring target clothing information according to the monitoring target clothing information and the preset weight value of the monitoring target clothing information;

Judging the clothing state of the target to be monitored according to the preset parameter threshold and the fourth confidence level of the clothing information of the monitoring target.

Further, judging whether there is an abnormality in the target to be monitored according to the state of the target to be monitored includes:

If the action state of the target to be monitored is fighting, there is an abnormality in the target to be monitored;

If the stay status of the target to be monitored is too long, there is an abnormality in the target to be monitored;

If the smoking status of the target to be monitored is smoking, the target to be monitored is abnormal;

If the clothing status of the target to be monitored is abnormal, the target to be monitored is abnormal

In the second aspect, the embodiment of the present invention also provides a safety monitoring system based on deep learning, including:

The first acquisition module is used to acquire the video data of the target to be monitored in the monitoring area;

The second acquisition module is used to analyze the video data through the pre-trained first model, so as to acquire the monitoring feature information of the target to be monitored in the monitoring area;

A judging module, configured to judge whether there is an abnormality in the target to be monitored according to the monitoring feature information;

The alarm module is used for performing an alarm operation when there is an abnormality in the target to be monitored.

In a third aspect, an embodiment of the present invention also provides a server, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the computer program, the steps of any one of the deep learning-based safety monitoring methods in the above embodiments are implemented.

In a fourth aspect, an embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of any one of the deep learning-based safety monitoring methods in the above-mentioned embodiments are implemented.

The present invention solves the technical problem in the prior art that the unstructured data form and content in the video cannot be analyzed and sorted out in time for a large amount of video information by performing deep learning-based training on the video data, resulting in data loss, and realizes the technical effects of semantically structured extraction of video data and effective management of video data.

Description of drawings

FIG. 1 is a flow chart of a safety monitoring method based on deep learning provided by Embodiment 1 of the present invention;

FIG. 2 is a flowchart of a safety monitoring method based on deep learning provided by Embodiment 2 of the present invention;

FIG. 3 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention;

FIG. 4 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention;

FIG. 5 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention;

FIG. 6 is a flow chart of another deep learning-based safety monitoring method provided in Embodiment 2 of the present invention;

FIG. 7 is a schematic structural diagram of a safety monitoring system based on deep learning provided by Embodiment 3 of the present invention;

FIG. 8 is a schematic structural diagram of a server provided by Embodiment 4 of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, only some structures related to the present invention are shown in the drawings but not all structures.

Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processing, many of the steps may be performed in parallel, concurrently, or simultaneously. Additionally, the order of steps may be rearranged. A process may be terminated when its operations are complete, but may also have additional steps not included in the figure. A process may correspond to a method, function, procedure, subroutine, subroutine, or the like.

In addition, the terms "first", "second", etc. may be used herein to describe various directions, actions, steps or elements, etc., but these directions, actions, steps or elements are not limited by these terms. These terms are only used to distinguish a first direction, action, step or element from another direction, action, step or element. For example, a first acquisition module could be termed a second acquisition module, and, similarly, a second acquisition module could be termed a first acquisition module, without departing from the scope of the present application. Both the first acquisition module and the second acquisition module are acquisition modules, but they are not the same acquisition module. The terms "first", "second", etc. should not be interpreted as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature defined as "first" and "second" may explicitly or implicitly include one or more of these features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

Embodiment one

Fig. 1 is a flow chart of a safety monitoring method based on deep learning provided by Embodiment 1 of the present invention. This embodiment is applicable to video analysis of targets to be monitored in a monitoring area, and the method can be executed by a processor. As shown in Figure 1, the safety monitoring method based on deep learning specifically includes the following steps:

Step S110, acquiring the video data of the target to be monitored in the monitoring area;

Specifically, monitoring personnel can install monitoring equipment such as monitoring cameras in areas that need to be monitored, monitor the monitoring areas in real time, and obtain corresponding monitoring videos.

Step S120, analyzing the video data through the pre-trained first model to obtain monitoring feature information of the target to be monitored in the monitoring area;

Specifically, the monitoring personnel can pre-establish a training model for data analysis of the video data. After obtaining the video data, the monitoring personnel can analyze the video data through the pre-established training model, thereby identifying or extracting the data contained in the video content (such as the behavior data of the target to be monitored), that is, the monitoring feature information.

Step S130, judging whether the target to be monitored is abnormal according to the monitoring feature information;

Specifically, after the detection feature information is obtained by training the model in step S120, it is judged according to the monitoring feature information whether there is an abnormality in the target to be monitored. In this embodiment, the monitoring feature information may include one or more of monitoring target action information, monitoring target dwell time information, smoke detection information, and monitoring target clothing information.

Step S140, performing an alarm operation when there is an abnormality in the target to be monitored.

Specifically, it is possible to judge whether the target to be monitored is abnormal according to the threshold or confidence of the monitoring feature information. For example, when the monitoring feature information is the action information of the monitoring target, if the confidence of the monitoring target’s action information is greater than or equal to the preset confidence threshold, it means that there is no abnormality in the monitoring target’s action information;

The beneficial effect of Embodiment 1 of the present invention is that by performing deep learning-based training on video data, it solves the technical problem in the prior art that the unstructured data form and content in the video cannot be analyzed and sorted out in a timely manner, resulting in data loss, and achieves the technical effects of semantically structured extraction of video data and effective management of video data.

Embodiment two

Embodiment 2 of the present invention is further optimized on the basis of Embodiment 1. FIG. 2 is a flow chart of a safety monitoring method based on deep learning according to Embodiment 2 of the present invention. As shown in Figure 2, the safety monitoring method based on deep learning of the present embodiment includes:

Step S210, acquiring the video data of the target to be monitored in the monitoring area;

Specifically, monitoring personnel can install monitoring equipment such as monitoring cameras in areas that need to be monitored, monitor the monitoring areas in real time, and obtain corresponding monitoring videos.

In this embodiment, the monitoring feature information may include one or more of monitoring target action information, monitoring target dwell time information, smoke detection information, and monitoring target clothing information.

Step S220, training the convolutional neural network on the video data to obtain a first model corresponding to the analysis of the monitoring target's action information, the monitoring target's stay time information, smoke detection information and the monitoring target's clothing information;

Specifically, monitoring personnel or staff can pre-set a training model for analyzing video data. For example, continuous monitoring and tracking of people, vehicles, and non-motor vehicles in surveillance video images, and optimal selection of key frame images for attribute recognition of people and vehicles, so as to obtain monitoring feature information such as license plates, models, brands, genders, ages, and clothing. In this embodiment, after acquiring structured information such as monitoring target action information, monitoring target duration information, smoke detection information, and monitoring target clothing information monitoring feature information, monitoring personnel or staff can realize the storage, calculation and application of structured video data through large-scale parallel processing databases, data mining, distributed file systems, sub-written test databases, and cloud computing platform scalable storage systems. The obtained data information is stored in a corresponding storage system in a structured form, so as to realize effective analysis, organization and management of video data.

Step S231, using the first model to identify the video data based on the monitoring target action information, so as to obtain the monitoring target action information of the target to be monitored;

Specifically, in this embodiment, human body behavior recognition may use skeleton behavior detection, that is, joint point estimation (Pose Estimation) is performed through redgreen blue (RGB) images. Each moment (frame) skeleton corresponds to the coordinate position information of the 18 joint points of the human body. A time sequence consists of several frames. Behavior recognition is to determine the type of behavior that belongs to the sequence that is pre-segmented in the time domain, that is, "reading behavior". To identify the skeletons of people in a specific area, especially for the identification of whether people in the monitoring area are fighting or not, the whole process has gone through the initial approach stage, the climax stage of waving fists and feet, and the end stage. In contrast, the climax stage of swinging fists and feet contains more information, which is most helpful for action discrimination. According to the temporal attention model, a sub-network is used to automatically learn and learn the importance of different frames in the sequence, so that important frames play a greater role in classification to optimize the accuracy of recognition.

Step S232, calculating the first confidence level of the monitoring target action information according to the first preset weight value;

Specifically, in this embodiment, the monitoring of the movement of the target to be monitored mainly depends on the body movement of the human body and the judgment of the relative position and moving speed. When calculating the first degree of confidence, the optimized optical flow method (optical flow method refers to a simple and practical expression of image motion, usually defined as the apparent motion of the image brightness pattern in an image sequence, that is, the expression of the motion speed of points on the surface of the space object on the imaging plane of the visual sensor), can be used to calculate the relative position, motion vector, relative motion speed and the speed of limb contact of the human body. Confidence, also known as reliability, or confidence level, confidence coefficient, that is, when sampling estimates the overall parameters, the conclusion is always uncertain due to the randomness of the sample. Therefore, a probability statement method is adopted, that is, the interval estimation method in mathematical statistics, that is, the estimated value and the overall parameter are within a certain allowable error range, and the corresponding probability is called confidence or confidence.

Step S233, judging the action state of the target to be monitored according to the preset parameter threshold and the first confidence level of the action information of the monitored target.

Specifically, the parameter threshold of the algorithm can be set in advance, and corresponding weight values can be set for different parameter thresholds. After calculating the first confidence level of the monitoring target action information according to different weight values, parameter thresholds and monitoring target action information, the action state of the target to be monitored can be judged according to the first confidence level. For example, when the monitoring personnel or staff need to identify fighting behavior, if the first confidence level is greater than or equal to the confidence level threshold (which can be the preset parameter threshold value of the monitoring target’s action information), it means that there is no abnormality in the monitoring area, that is, there is no abnormality in the action state of the target to be monitored, that is to say, there is no fight; Monitoring personnel or staff conduct safety inspections on the monitoring area to ensure the safety of the monitoring area. In this embodiment, facial expression recognition can also be assisted, so as to judge the action state of the target to be monitored more accurately.

FIG. 3 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention. FIG. 4 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention. FIG. 5 is a flow chart of a safety monitoring method based on deep learning in an alternative embodiment provided by Embodiment 2 of the present invention.

As shown in Figure 3, an alternative embodiment of step S231 to step S233 may be:

Step S241, using the first model to identify the video data based on the duration information of the monitoring target, so as to obtain the duration information of the monitoring target of the target to be monitored;

Step S242, calculating the second confidence level of the monitoring target staying time information according to the second preset weight value and the monitoring target staying time information;

Step S243, judging the stay state of the target to be monitored according to the preset parameter threshold and the second confidence level of the monitoring target's stay duration information.

Specifically, in this embodiment, monitoring may be performed on the residence time of the target to be monitored in the monitoring area. In the first model of monitoring target staying time, the second confidence level is calculated according to the second preset weight value and the monitoring target staying time information, and then the staying state of the target to be monitored is judged according to the second confidence level, the preset parameter threshold and the second confidence level. When the second confidence level is greater than or equal to the preset confidence level threshold (it can be the preset parameter threshold value of the monitoring target’s stay duration information), it means that the stay state of the target to be monitored is a normal stop, and when the second confidence level is less than the preset confidence level threshold, it means that the stay state of the target to be monitored is too long.

As shown in Figure 4, the alternative embodiment of step S231 to step S233 can also be:

Step S251, using the first model to identify the video data based on the smoke detection information, so as to obtain the smoke detection information of the target to be monitored;

Step S252, calculating the smoke concentration according to the smoke detection information and determining the area where the smoke is located and the smoking action of the target to be monitored;

Step S253, calculating the third confidence level of the smoke detection information according to the smoke concentration, the area where the smoke is located, and the smoking action of the target to be monitored;

Step S254, judging the smoking state of the target to be monitored according to the preset parameter threshold of the smoke detection information and the third confidence level.

Specifically, in this embodiment, the smoking state of the target to be monitored can be detected. Smoking detection mainly depends on the detection of smoke concentration, the detection of smoke location and the detection of smoking action. Therefore, three neural networks are required to detect the smoke concentration, the area where the smoke is located, and the smoking action of the target to be monitored, and the third confidence degree of the smoke concentration, the area where the smoke is located, and the smoking action are respectively obtained under the corresponding neural network. The third confidence degree of the three is calculated according to different weights to obtain the third confidence degree of the smoke detection information. When the third confidence level is greater than or equal to the confidence level threshold (which can be the preset parameter threshold value of the smoke detection information), it means that the smoking state of the target to be monitored is non-smoking, and when the third confidence level is less than the confidence level threshold, it means that the smoking state of the target to be monitored is smoking. At this time, the safety monitoring system will send an alarm signal, thereby notifying the monitoring personnel or staff to check the personnel in the monitoring area to ensure the health of the air in the monitoring area and the safety of the facilities.

As shown in Figure 5, an alternative embodiment of step S231 to step S233 may also be:

Step S261, using the first model to identify the video data based on the monitoring target clothing information, so as to obtain the monitoring target clothing information of the target to be monitored;

Step S262, calculating the fourth confidence level of the monitoring target clothing information according to the monitoring target clothing information and the preset weight value of the monitoring target clothing information;

Step S263, judging the clothing state of the target to be monitored according to the preset parameter threshold and the fourth confidence level of the clothing information of the monitoring target.

Specifically, in this embodiment, it is possible to detect the clothing state of the target to be monitored. When detecting the state of clothing, the clothing is mainly identified through the color of the clothing and the special logo of the clothing, and the detection of the clothing color needs to consider the error caused by the environment and color difference. Therefore, before the detection of the dressing state of the target to be monitored, the neural network can be used to identify the data sets of different clothing colors and the data sets of the special logo of different clothing. After obtaining the monitoring target clothing information of the target to be monitored, the fourth confidence degree can be calculated according to different weight values. When the fourth confidence degree is greater than or equal to the confidence threshold (which can be the preset parameter threshold of the monitoring target clothing information), it means that the clothing status of the target to be monitored is normal clothing, and when the fourth confidence degree is less than the confidence threshold, it indicates that the clothing status of the target to be monitored is abnormal clothing. At this time, the safety monitoring system will send an alarm signal, thereby notifying the monitoring personnel or staff to check the personnel in the monitoring area to prevent suspicious personnel from endangering the safety of others or public facilities.

FIG. 7 is a schematic structural diagram of a safety monitoring system based on deep learning provided by Embodiment 3 of the present invention. In this embodiment, the safety monitoring method based on deep learning also includes:

Step S271, if the action state of the target to be monitored is fighting, then there is an abnormality in the target to be monitored;

Step S272, if the stay status of the target to be monitored is too long, then there is an abnormality in the target to be monitored;

Step S273, if the smoking state of the target to be monitored is smoking, the target to be monitored is abnormal;

Step S274, if the clothing status of the target to be monitored is abnormal, then the target to be monitored is abnormal;

Step S280, performing an alarm operation when there is an abnormality in the target to be monitored.

Specifically, the action state, lingering state, smoking state, and dressing state of the target to be monitored can be judged according to the first, second, third, and fourth confidence levels, respectively. When the status of any target to be monitored is abnormal, the safety monitoring system will perform an alarm operation, thereby notifying the monitoring personnel or security personnel of possible safety hazards in the monitoring area, and then take targeted preventive measures.

The beneficial effect of Embodiment 2 of the present invention is that by performing deep learning-based training on video data, and performing data analysis in the corresponding neural network for the information of the target to be monitored in the video data, it solves the technical problem in the prior art that the unstructured data form and content in the video cannot be analyzed and sorted out in time for a large amount of video information, resulting in data loss, and realizes the technical effect of semantically structured extraction of video data and effective management of video data.

Embodiment three

FIG. 3 is a schematic structural diagram of a safety monitoring system based on deep learning provided by Embodiment 3 of the present invention. As shown in Figure 3, the safety monitoring system 300 based on deep learning of this embodiment includes:

The first obtaining module 310 is used to obtain the video data of the target to be monitored in the monitoring area;

The second acquiring module 320 is configured to analyze the video data through the pre-trained first model, so as to acquire the monitoring characteristic information of the target to be monitored in the monitoring area;

Judging module 330, configured to judge whether there is an abnormality in the target to be monitored according to the monitoring feature information;

The alarm module 340 is configured to perform an alarm operation when there is an abnormality in the target to be monitored.

In this embodiment, the monitoring feature information includes one or more of monitoring target action information, monitoring target stay time information, smoke detection information, and monitoring target clothing information.

In this embodiment, the safety monitoring system 300 based on deep learning also includes:

The training module 350 is used to train the convolutional neural network on the video data to obtain a first model corresponding to the analysis of the monitored target's action information, the monitored target's stay time information, the smoke detection information and the monitored target's clothing information.

In this embodiment, the second acquisition module 320 includes:

The first training unit is configured to use the first model to identify the video data based on the monitoring target action information, so as to obtain the monitoring target action information of the target to be monitored;

A first calculation unit, configured to calculate a first confidence level of the monitoring target action information according to a first preset weight value;

The first judging unit is configured to judge the action state of the target to be monitored according to the preset parameter threshold and the first confidence level of the action information of the monitored target.

In this embodiment, the second acquiring module 320 also includes:

The second training unit is used to use the first model to identify the video data based on the duration information of the monitoring target, so as to obtain the duration information of the monitoring target of the target to be monitored;

The second calculation unit is used to calculate the second confidence level of the monitoring target staying time information according to the second preset weight value and the monitoring target staying time information;

The second judging unit is configured to judge the stay status of the target to be monitored according to the preset parameter threshold and the second confidence level of the stay duration information of the monitored target.

In this embodiment, the second acquiring module 320 also includes:

The third training unit is used to use the first model to identify the video data based on the smoke detection information, so as to obtain the smoke detection information of the target to be monitored;

The third calculation unit is used to calculate the smoke concentration according to the smoke detection information and determine the area where the smoke is located and the smoking action of the target to be monitored;

The fourth calculation unit is used to calculate the third confidence level of the smoke detection information according to the smoke concentration, the area where the smoke is located, and the smoking action of the target to be monitored;

The third judging unit is configured to judge the smoking state of the target to be monitored according to the preset parameter threshold and the third confidence level of the smoke detection information.

In this embodiment, the second acquiring module 320 also includes:

The fourth training unit is used to use the first model to identify the video data based on the monitoring target clothing information, so as to obtain the monitoring target clothing information of the target to be monitored;

The fifth calculation unit is used to calculate the fourth confidence degree of the monitoring target clothing information according to the monitoring target clothing information and the preset weight value of the monitoring target clothing information;

The fourth judging unit is used for judging the clothing state of the target to be monitored according to the preset parameter threshold and the fourth confidence level of the clothing information of the monitoring target.

In this embodiment, the judging module 330 includes:

The fifth judging unit is configured to determine that the target to be monitored is abnormal if the action state of the target to be monitored is fighting;

The sixth judging unit is used to determine that the target to be monitored is abnormal if the stay state of the target to be monitored is too long;

The seventh judging unit is used to determine that the target to be monitored is abnormal if the smoking state of the target to be monitored is smoking;

The eighth judging unit is configured to determine that the target to be monitored is abnormal if the clothing state of the target to be monitored is abnormal.

The deep learning-based safety monitoring system provided by the embodiments of the present invention can execute the deep learning-based safety monitoring method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

Embodiment four

4 is a schematic structural diagram of a server provided in Embodiment 4 of the present invention. As shown in FIG. 4 , the server includes a processor 410, a memory 420, an input device 430, and an output device 440; the number of processors 410 in the server can be one or more, and one processor 410 is used as an example in FIG.

The memory 410, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions/modules corresponding to the safety monitoring system based on deep learning in the embodiment of the present invention (for example, the first acquisition module, the second acquisition module, the judgment module, the alarm module and the training module of the safety monitoring system based on deep learning). The processor 410 executes various functional applications and data processing of the server by running the software programs, instructions and modules stored in the memory 420, that is, implements the above-mentioned security monitoring method based on deep learning.

That is:

Obtain the video data of the target to be monitored in the monitoring area;

Analyzing the video data by using the pre-trained first model to obtain monitoring feature information of the target to be monitored in the monitoring area;

Judging whether the target to be monitored is abnormal according to the monitoring characteristic information;

When there is an abnormality in the target to be monitored, an alarm operation is performed.

The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application program required by at least one function; the data storage area may store data created according to the use of the terminal, and the like. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage devices. In some examples, the memory 420 may further include a memory that is remotely located relative to the processor 410, and these remote memories may be connected to a server through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The input device 430 can be used to receive input numbers or character information, and generate key signal input related to user settings and function control of the server. The output device 440 may include a display device such as a display screen.

Embodiment five

Embodiment 5 of the present invention also provides a storage medium containing computer-executable instructions. When executed by a computer processor, the computer-executable instructions are used to execute a safety monitoring method based on deep learning. The method includes:

Obtain the video data of the target to be monitored in the monitoring area;

Analyzing the video data by using the pre-trained first model to obtain monitoring feature information of the target to be monitored in the monitoring area;

Judging whether the target to be monitored is abnormal according to the monitoring characteristic information;

When there is an abnormality in the target to be monitored, an alarm operation is performed.

Certainly, a storage medium containing computer-executable instructions provided by an embodiment of the present invention, the computer-executable instructions are not limited to the operations of the above method, and may also perform related operations in the deep learning-based safety monitoring method provided by any embodiment of the present invention.

Through the above description of the implementation, those skilled in the art can clearly understand that the present invention can be implemented by software and necessary general hardware, and of course also by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (FLASH), a hard disk or an optical disk, etc., including several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the various embodiments of the present invention. method.

It is worth noting that in the above-mentioned embodiment of the safety monitoring system based on deep learning, the units and modules included are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the present invention.

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention, and the present invention The scope is determined by the scope of the appended claims.

Claims (8)

1. A safety monitoring method based on deep learning, characterized in that, comprising: Obtain the video data of the target to be monitored in the monitoring area; Analyzing the video data by using a pre-trained first model to obtain monitoring feature information of the target to be monitored in the monitoring area; judging whether there is an abnormality in the target to be monitored according to the monitoring characteristic information; Carrying out an alarm operation when there is an abnormality in the target to be monitored; Wherein, the monitoring feature information includes one or more of monitoring target action information, monitoring target stay time information, smoke detection information and monitoring target clothing information; Wherein, before analyzing the video data through the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area, the steps include: Carrying out convolutional neural network training on the video data to obtain a first model corresponding to the monitoring target action information, monitoring target stay time information, smoke detection information and monitoring target clothing information; Wherein, analyzing the video data through the pre-trained first model to obtain the monitoring characteristic information of the target to be monitored in the monitoring area includes: Using the first model to identify the video data based on the duration information of the monitoring target, so as to obtain the duration information of the monitoring target of the target to be monitored; calculating a second confidence level of the monitoring target stay information according to a second preset weight value and the monitoring target stay information; judging the stay state of the target to be monitored according to the preset parameter threshold of the stay duration information of the monitoring target and the second confidence level; Wherein, the stay time information of the monitoring target is identified and determined by monitoring and tracking the monitor video images, and selecting key frame images preferably. 2. A kind of safety monitoring method based on deep learning according to claim 1, characterized in that, the analysis of the video data by the pre-trained first model to obtain the monitoring feature information of the target to be monitored in the monitoring area comprises: Using the first model to identify the video data based on the monitoring target action information, so as to obtain the monitoring target action information of the target to be monitored; calculating a first confidence level of the monitoring target action information according to a first preset weight value; The action state of the target to be monitored is judged according to the preset parameter threshold of the action information of the monitored target and the first confidence level. 3. A kind of security monitoring method based on deep learning according to claim 1, characterized in that, analyzing the video data through the first model trained in advance to obtain the monitoring characteristic information of the target to be monitored in the monitoring area also includes: Using the first model to identify the video data based on the smoke detection information, so as to obtain the smoke detection information of the target to be monitored; calculating the smoke concentration according to the smoke detection information and determining the area where the smoke is located and the smoking action of the target to be monitored; calculating a third confidence level of the smoke detection information according to the smoke concentration, the area where the smoke is located, and the smoking action of the target to be monitored; judging the smoking state of the target to be monitored according to the preset parameter threshold of the smoke detection information and the third confidence level; Wherein, the confidence levels of the smoke concentration, the area where the smoke is located, and the smoking action of the target to be monitored are determined by detecting the concentration of smoke, the area where the smoke is located, and the smoking action of the target to be monitored through different neural networks. 4. A kind of safety monitoring method based on deep learning according to claim 1, it is characterized in that, the described video data is analyzed by the first model trained in advance, to obtain the monitoring feature information of the target to be monitored in the monitoring area also includes: Using the first model to identify the video data based on the monitoring target clothing information, so as to obtain the monitoring target clothing information of the target to be monitored; calculating the fourth confidence level of the monitoring target clothing information according to the monitoring target clothing information and the preset weight value of the monitoring target clothing information; Judging the clothing state of the target to be monitored according to the preset parameter threshold of the clothing information of the monitoring target and the fourth confidence level. 5. A safety monitoring method based on deep learning according to claim 1-4, wherein the judging whether there is an abnormality in the target to be monitored according to the state of the target to be monitored comprises: If the action state of the target to be monitored is fighting, the target to be monitored is abnormal; If the stay status of the target to be monitored is too long, the target to be monitored is abnormal; If the smoking state of the target to be monitored is smoking, the target to be monitored is abnormal; If the clothing status of the target to be monitored is abnormal, then the target to be monitored is abnormal. 6. A safety monitoring system based on deep learning, characterized in that, comprising: The first acquisition module is used to acquire the video data of the target to be monitored in the monitoring area; The second acquisition module is configured to analyze the video data through the pre-trained first model, so as to acquire the monitoring feature information of the target to be monitored in the monitoring area; A judging module, configured to judge whether there is an abnormality in the target to be monitored according to the monitoring characteristic information; An alarm module, configured to perform an alarm operation when the target to be monitored is abnormal; Wherein, the monitoring feature information includes one or more of monitoring target action information, monitoring target stay time information, smoke detection information and monitoring target clothing information; Wherein, the training module is used to perform convolutional neural network training on the video data, so as to obtain a first model corresponding to the monitoring target action information, monitoring target stay time information, smoke detection information and monitoring target clothing information; Wherein, the first training unit is configured to use the first model to identify the video data based on the duration information of the monitoring target, so as to obtain the duration information of the monitoring target of the target to be monitored; A first calculation unit, configured to calculate a second confidence level of the monitoring target stay information according to a second preset weight value and the monitoring target stay information; A first judging unit, configured to judge the stay status of the target to be monitored according to the preset parameter threshold of the stay duration information of the monitoring target and the second confidence level; Wherein, the stay time information of the monitoring target is identified and determined by monitoring and tracking the monitor video images, and selecting key frame images preferably. 7. A server, comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein the processor implements the steps of the deep learning-based safety monitoring method according to any one of claims 1-5 when executing the computer program. 8. A computer-readable storage medium, on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of the deep learning-based safety monitoring method according to any one of claims 1-5 are implemented.