TWI658789B - Automatic fishery and preys identification and recording system - Google Patents

Abstract

The present invention provides a device for recognizing and recording a catch, comprising: a camera device, a programmable video device, a recognition detecting unit and a memory storage unit. The programmable video recording device is configured to capture a dynamic image through a first depth learning training to identify whether the dynamic image presents the occurrence of the catch, and activate the camera device to obtain the one of the catches due to the occurrence of the catch. The first fish image information; the identification detecting unit is electrically connected to the programmable video device, and the first fish image information is obtained from the programmable video device, and the fish body is obtained through a second deep learning training. a first identification result; the memory storage unit is electrically connected to the programmable video device and the camera device, and configured to store the motion image information of the motion image and the first fish image information.

Description

Catch and prey automatic identification login system

The present invention relates to an identification system; in particular, to a catch and prey automatic identification login system.

Due to the active fishing in the past half century and the depletion of marine resources, countries around the world have begun to pay attention to the statistics of catches in various regions each year, especially the capture of large-scale fish such as squid, shark or swordfish at the top of the marine bio-chain. Collection and analysis of records. The offshore fishing industry in Taiwan is very developed. Taiwanese fishing boats are spread all over the fish farms in the oceans. However, the current record collection methods for these large catches are still in the traditional manual recording stage, and it is difficult to provide international requirements for these catch statistics.

In order to be able to provide international fishery unit data for accurate and detailed global fishery resource statistics, the entry of large catches such as catfish, sharks or swordfish caught by fishing vessels must be immediately reflected in the time and place of capture and the type and size of catch. According to the current traditional manual recording method, it is difficult for Taiwan's fishery units to provide persuasive real-time catch data to the international community, which easily constitutes a misunderstanding between each other. Currently used in some countries A machine observer system is installed on the ocean fishing vessels to record the catch data through dynamic photography or photographing. However, for the identification of catch species, it still depends on manual work.

Therefore, how to simultaneously record the catch automatically and record the catch information and the navigation data of the fishing vessel. The records include information on the type of catch, body length, weight and information on fishing vessels. These catch and fishing vessel information can be synchronized to the shore, to achieve real-time data updates, to achieve the purpose of fishery data statistics, is a technical problem that needs to be resolved. Similar biometric needs are also present in the application of wildlife hunting records on the road.

Therefore, there is a need to develop a device or system that can instantly identify and record catches without the intervention of manual operations, and can provide the required identification and other basic information in a timely and correct manner through the assistance of the automated electronic device.

According to one aspect of the present invention, a system for recognizing and recording a catch is provided, comprising: a camera device, a programmable video device, a recognition detection unit, and a memory storage unit. The programmable video recording device is configured to capture a dynamic image, and use a first depth learning training to identify whether the dynamic image presents the occurrence of the catch, and activate the camera device to obtain the catch due to the occurrence of the catch. a first fish image information; the identification detecting unit is electrically connected to the programmable video device, and the first fish image information is obtained from the camera device and is learned by a second depth Training to obtain a first identification result of the fish body; the memory storage unit is electrically connected to the programmable video device and the camera device, and configured to store the motion image information of the motion image and the first fish image information .

According to another aspect of the present invention, an apparatus for instantly identifying a prey includes a camera device, a programmable video device, and a recognition detecting unit. The programmable video device is configured to continuously capture a motion image, and use a first depth learning training to identify whether the motion image presents the presence of the prey, and activate the camera device to obtain the prey due to the presence of the prey. a first biometric image information; the identification detecting unit is electrically connected to the programmable video recording device, obtains the first biometric image information from the camera device, and obtains one of the prey through a second deep learning training A recognition result.

The device utilizes a device with automatic identification capability to identify fish catches or prey, and can be used to record the number and location of wild animals that are hunt, and has industrial applicability.

10‧‧‧Photographing device

20‧‧‧Programmable video device

30‧‧‧ID detection unit

40‧‧‧Memory storage unit

50‧‧‧Global Positioning Unit

60‧‧‧Satellite communication unit

70‧‧‧Sensor unit

100‧‧‧Automatic identification login system

200‧‧‧ fishing boats

220‧‧‧ lifting device

230‧‧ ‧ warehouse entrance

A‧‧‧catch

B‧‧ ‧ prey

This case can be further understood by the following detailed description of the drawings: Figure 1 is a schematic diagram of an embodiment of the automatic identification and registration system for catch and prey of the present invention; and Figure 2 is an implementation of the automatic identification and registration system for the catch of the present invention. Schematic diagram Fig. 3 is a schematic view showing an embodiment of the prey automatic identification registration system of the present invention.

The present invention will be fully understood from the following description of the embodiments of the present invention, which can be practiced by those skilled in the art.

Deep learning is a method based on the use of machine learning to characterize learning in the field of information science. It has been used in biometrics in recent years. Computer programs have the ability to recognize specific organisms through deep learning from hundreds or even thousands of images with known specific biological images. The present invention utilizes different depth learning to obtain the most efficient instant automatic biometric identification.

Please refer to Fig. 1, which shows an embodiment of the automatic catch and prey automatic registration system of the present invention. The automatic identification registration system 100 shown in the drawing comprises a camera device 10, a programmable video recording device 20, a recognition detecting unit 30 and a memory storage unit 40. In order to provide surface location information (such as latitude and longitude) and corresponding time information in real time, the automatic identification login system 100 further includes a global positioning unit 50 and a satellite communication unit 60 electrically connected to the programmable video device 20.

The programmable video device 20 is configured to capture a motion picture (not shown), for example, to continuously observe video recordings of a specific area, usually 24 hours a day, so that manual manipulation can be avoided. Any hunting or fishing record. The program (not shown) carried in the programmable video device 20 has the ability to process and analyze images, and the first depth learning training is used to identify whether the motion image exhibits any catch or prey. Those skilled in the art can understand that the program of the programmable video device can be trained in the known image, that is, the first deep learning training, and has the ability to process and analyze the image from the recorded motion image. It is possible to interpret the images of the catches or prey that are being tracked, and there may be more than one type of catch or prey that can be tracked. The programmable video device 20 is electrically coupled to the camera device 10. Once it is determined that the captured catch or prey image may be present in the motion image, the programmable video device 20 activates the camera device 10 due to the occurrence of a catch or prey image. The camera 10 obtains high-resolution biological image information, such as fish images or prey images. According to an embodiment of the invention, the biometric information may be a group of photos taken continuously, which avoids the chance of missing any better shooting opportunities and angles.

The identification detecting unit 30 is electrically connected to the programmable video device 20, and the high-resolution biological image information (such as fish image or prey image) captured by the camera device 10 can be obtained from the programmable video device 20, according to these high The biometric information of the resolution is usually a set of single photos, and the identification detecting unit 30 obtains the identification result of the fish body or the prey through the training of the second deep learning. According to an embodiment of the invention, the camera device 10 has a high resolution continuous photography capability, preferably with automatic multifocal Capabilities, the above-mentioned set of single photos are usually multiple high-resolution photos taken continuously. Different from the first deep learning, the second deep learning function is to train the identification detecting unit 30 to be able to distinguish the fish image or the species to which the prey image belongs, such as large fish such as squid, shark or swordfish or sambar, Wild animals such as sika deer or hawthorn. For the interpretation of a group of multiple photos, there may be different species interpretation results. According to an embodiment of the invention, the maximum number of interpretation results may be selected as the identification result. In addition, the trained identification detection unit 30 via the second depth learning can also be used to identify the size of the fish or prey.

The memory storage unit 40 is electrically connected to the programmable video devices 20 and 10, and is configured to store dynamic image information of the dynamic image and high-resolution biological image information. In addition to being used for instant identification, these image information will also provide future verification and analysis to ensure that all image records are complete during the hunting period without any omissions.

The global positioning unit 50 can instantly provide time and surface location information when the catch or prey image appears, most commonly using latitude and longitude information; the satellite communication unit 60 is configured to instantly transmit the identification result of the fish or prey, the location information of the surface and corresponding The time information.

The programmable video device 20 can continuously perform video recording to avoid any missing pictures. However, in order to save video resources, the sensing unit 70 may be additionally configured in the automatic identification login system 100 to be electrically connected to the programmable video device 20, when the sensing unit 70 is subjected to catch or prey. The presence, for example, by means of sensing by means of weight or light interference, sends a signal to the programmable video device 20 to trigger activation of the programmable video device 20. Such a configuration allows the programmable video device 20 to not remain in a powered-on state. According to an embodiment, the sensing unit 70 can also be configured to trigger the camera device 10.

What is captured by the video device is actually a series of individual images. In accordance with another embodiment of the present invention, the programmable video device 20 is further configured to retrieve individual fish or biometric information from the motion image information. This task can be performed by a motion image capture program. When the program is trained in deep learning and has the ability to identify specific species, these individual fish or biometric information can be used to obtain identification of the fish or biological image, including species, body length and even weight. Estimate. In order to distinguish from the previous deep learning mode, this method of training the deep learning mode of the programmable video device 20 through the images captured from the dynamic image information is referred to herein as third depth learning.

Please refer to FIG. 2, which shows a schematic diagram of an embodiment of the automatic catch identification registration system of the present invention. The ocean-going fishing vessel 200 can be loaded with the automatic identification registration system 100 provided by the present invention. As shown in the figure, when a catch A is caught on the ship and the catch A is transported by the lifting device 220 to the warehouse entrance 230 on the deck, the configuration is configured. The programmable video device 20 and the camera device 10 on the deck of the fishing boat 200 can take pictures or take pictures in real time, or send signals according to the weight sensed by the sensing unit 70 on the lifting device 220. (Not shown) Activate photography or photo taking. The remaining devices in the automatic identification login system 100 can be loaded on other locations of the fishing boat 200 and need not be displayed in the figure. From the position of the programmable video device 20 and the camera device 10 shown in the figure, since the programmable video device 20 can monitor the entry and exit of any catch at the warehouse entrance 230 24 hours a day, it is possible to avoid missing any amount of fish in the warehouse. The position of the camera device 10 can be different from that of the programmable video device 20. Both can capture the catch A from different shooting angles at the same time, and the captured image data can be used for mutual comparison.

Please refer to FIG. 3, which shows a schematic diagram of an embodiment of the prey automatic identification registration system of the present invention. In addition to the fishing vessel at sea, the automatic identification registration system 100 provided by the present invention can be loaded to obtain and transmit fish identification results for catch statistics. The automatic identification registration system 100 provided by the present invention can also be applied to biometric identification on land. recording. The sensing unit 70 disposed in the woods in the figure may be a device that uses optical sensing to determine whether the prey B is present, for example, an optical reflecting device 71 is disposed at a corresponding position, so that the sensing unit 70 can be opposite to the optical reflecting device 71. The optical signal (not shown) is continuously emitted and receives an optical signal (not shown) from the optical reflecting device 71. When the prey B passes the position where the optical signal indicated by the dotted arrow passes, the optical signal is blocked, so the sensing unit 70 can send a trigger signal to activate the programmable video device 20 or the camera 10.

In an embodiment of the invention, the programmable video device 20 The program can be pre-processed by known images, that is, the first deep learning training described above, and thus has the ability to process and analyze images, and whether the tracked prey images may appear in the recorded motion images. . The programmable video device 20 and the camera device 10 can be connected via a wired or wireless connection. Once it is determined that the tracked prey image may appear in the motion image, the programmable video device 20 activates the camera device 10 due to the presence of the prey image. The photographic apparatus 10 obtains high-resolution bio-image information. If, by analysis, it is considered that the prey appearing in the motion image does not conform to the characteristics of any tracked object, the programmable video device 20 does not activate the camera device 10, which can prevent the camera device 10 from wasting memory due to blind triggering. space.

In another embodiment, the first depth learning training can enable the programmable video device 20 to recognize whether the motion image contains a suspected prey, and obtain the first biological image of the suspected prey in response to the confirmation of one of the suspected prey. Information, such as taking a set of single photos from the motion picture. Referring again to FIG. 1, the identification detecting unit 30 obtains the suspected prey as the specific prey through a second deep learning training. The identification detecting unit 30 can perform identification based on the first biometric image information from the programmable video device 20 or the second biometric image information obtained from the camera device 10. Generally, the resolution of the image information obtained by the camera device 10 is better, and it is easy to obtain a relatively accurate recognition result. However, when the camera device 10 in the field fails, the programmable video device 20 can still be used. The first biometric information is used to perform the identification and maintain the basic functions of the system.

Example

A system for recognizing and recording a catch, comprising: a camera device; a programmable video device configured to capture a motion image, and a first depth learning training to identify whether the motion image presents the capture Appearing and launching the camera device to obtain the first fish image information of the catch due to the occurrence of the catch; a recognition detecting unit electrically connected to the programmable video device, obtaining the first from the camera device Fish body image information, and obtaining a first identification result of a fish body through a second deep learning training; and a memory storage unit electrically connected to the programmable video device and the camera device, and configured to store the Dynamic image information of the moving image and the first fish image information.

2. The device of embodiment 1, wherein the first identification result is at least one of a fish species and an integral length of the fish body.

3. The apparatus of embodiment 1, further comprising: a global positioning unit that provides a surface location information and a corresponding time information; and a satellite communication unit configured to immediately transmit the first identification result, the surface location Information and corresponding time information.

4. The device of embodiment 1, wherein the programmable video device is further configured to obtain a second fish image information from the motion image information and obtain the fish body through a third depth learning training. A second identification result.

5. The device of embodiment 1, further comprising: a sensing unit configured to instantly trigger the programmable video device.

6. A device for instantly recognizing a prey, comprising: a camera device; a programmable video device configured to continuously capture a motion image, and a first depth learning training to identify whether the motion image presents the prey Appearing, and starting the camera device to obtain the first biometric image information of the prey due to the presence of the prey; and an identification detecting unit electrically connected to the programmable video device, obtaining the first from the camera device The biometric information is obtained through a second deep learning training to obtain the first identification result of the prey.

7. The device of embodiment 6, wherein the first identification is at least one of a variety and an integral type of the prey.

8. The device of embodiment 6, further comprising: a memory storage unit electrically coupled to the programmable video device and the camera device, and configured to store motion image information of the motion image and the first biological image information a global positioning unit providing a surface location information and corresponding time information; and a satellite communication unit configured to immediately transmit the first identification result, the surface location information, and the corresponding time information.

9. The device of embodiment 6, wherein the programmable video device is further configured to obtain a second biometric image information from the motion image information, and obtain a prey of the prey through a third depth learning training. Two identification results.

10. The device of embodiment 6, further comprising: a sensing unit configured to instantly trigger the programmable video device.

11. A device for instantly identifying a particular prey comprising: A programmable video device configured to continuously capture a motion picture and identify whether the motion picture contains a suspected prey through a first depth learning training, and obtain the suspected prey in response to the confirmation of one of the suspected prey a first biometric information; and an identification detecting unit electrically connected to the programmable video device, obtaining the first biometric information, and training through a second deep learning to obtain whether the suspected prey is the specific prey One of the first identification results.

12. The apparatus of embodiment 11, further comprising: a camera device for providing a second biometric image information for the confirmation of the programmable video recording device, the identification detecting unit utilizing the second biometric image information Obtaining whether the suspected prey is the second identification result of one of the specific prey.

13. The device of embodiment 12 wherein the first biometric information is provided by the programmable video device.

The present invention has been described with reference to the preferred embodiments and examples thereof, which are intended to be illustrative and not restrictive. It will be understood by those skilled in the art that various combinations and modifications may be made without departing from the spirit and scope of the invention.

Claims (5)

A system for recognizing and recording a catch comprises: a camera device; a programmable video device configured to capture a motion image, and a first depth learning training to identify whether the motion image exhibits the occurrence of the catch And launching the camera device to obtain the first fish image information of the catch due to the occurrence of the catch, wherein the first fish image information is a set of multiple high-resolution photos; a recognition detection unit, electricity Connecting to the programmable video device, obtaining the first fish image information from the camera device, and obtaining a first identification result of a fish body through a second deep learning training, wherein the first identification result is about At least one of the fish species and the integrated length of the fish body, and the first identification result is the maximum number of interpretation results after the interpretation of the plurality of high-resolution photos of the group; and a memory storage unit electrically connected to the The programmable video device and the camera device are configured to store the motion image information of the motion image and the first fish image information. The system of claim 1, further comprising: a global positioning unit that provides a surface location information and corresponding time information; a satellite communication unit configured to immediately transmit the first identification result, the surface location Information and corresponding time information; and a sensing unit configured to instantly trigger the programmable video device. The system of claim 1, wherein the programmable video device is further configured to obtain a second fish image information from the motion image information, and obtain the fish body through a third depth learning training. One of the second identification results. A device for instantly recognizing a prey comprises: a camera device; a programmable video device configured to continuously capture a motion image, and a first depth learning training to identify whether the motion image exhibits the presence of the prey And starting the camera device to obtain the first biometric image information of the prey according to the presence of the prey, wherein the first biometric image information is a set of multiple high resolution photos; and an identification detection unit, electrical connection Obtaining the first biometric image information from the camera device, and obtaining a first identification result of the prey through a second depth learning training, wherein the first identification result is related to the prey At least one of a variety and a combination type, and the first identification result is a maximum number of interpretation results after the interpretation of the plurality of high-resolution photos of the group. The device of claim 4, further comprising: a memory storage unit electrically connected to the programmable video device and the camera device, and configured to store the motion image information of the motion image and the first biological image Information; a global positioning unit providing a surface location information and corresponding time information; a satellite communication unit configured to immediately transmit the first identification result, the surface location information and the corresponding time information; and a sensing unit , configured to instantly trigger the programmable video device.