CN111476816A

CN111476816A - Intelligent efficient simultaneous recognition method for multiple objects

Info

Publication number: CN111476816A
Application number: CN201910930816.7A
Authority: CN
Inventors: 罗顺发; 赵肖彬; 崔伟勋; 罗嗣达; 张辉
Original assignee: Jabsco Electronic Technology Co ltd
Current assignee: Jabsco Electronic Technology Co ltd
Priority date: 2019-09-29
Filing date: 2019-09-29
Publication date: 2020-07-31

Abstract

Compared with the prior identification technology, the method has the advantages that the target image and the environment image are independently marked after the operation according to the steps of the method, the integral comparison calculation of the integral picture is not needed, only the position change between the target image and the separated target point is calculated, the memory is saved, the CPU resource is saved, and the identification efficiency is higher.

Description

Intelligent efficient simultaneous recognition method for multiple objects

Technical Field

The invention relates to the field of object identification, in particular to an intelligent efficient simultaneous identification method for multiple objects.

Background

At present, along with the increasingly severe world safety situation, the lives of urban citizens are seriously threatened by wanted victims, escaped prisoners, stolen vehicles and the like in cities. People need to monitor these special types of targets and track and predict the moving tracks of the targets. This requires the identification of this type of object and the identification of the location of this type of object on a city map for the police to control it.

The existing identification technology judges whether to move to identify objects or not by learning a model and then comparing a plurality of frames of data with the model, and has the defects that: the data volume of a single model is large, the data volume of a multi-object model is several times of that of a single model, and CPU resources are consumed seriously. Therefore, those skilled in the art provide an intelligent method for simultaneously recognizing multiple objects with high efficiency to solve the above problems in the background art.

Disclosure of Invention

The invention aims to provide an intelligent method for simultaneously identifying multiple objects efficiently, so as to solve the problems in the background technology.

In order to achieve the purpose, the invention provides the following technical scheme:

an intelligent efficient simultaneous recognition method for multiple objects comprises the following steps:

1) obtaining a picture with a fixed visual angle through monitoring equipment, obtaining an environment image, and forming an index list of all coordinate information;

2) inputting a plurality of groups of comparison images, marking the approximate outline of the target object, and separating the approximate outline from the whole picture;

3) extracting target contour information, extracting characteristic points of the obtained image, and extracting and concentrating the characteristic points of the obtained image in the index list to obtain a characteristic point set C contained in a target contour;

4) forming a background reference coordinate set B of the moving path;

5) normalizing the pixel coordinates of the image to be recognized to obtain a pixel coordinate matrix of the image according to the moved image screenshot, and comparing the feature point set C with coordinate information in the image to obtain the position of the feature point set C in the image;

6) and obtaining the moving path of the feature point set C.

As a further scheme of the invention: in the step 1), the gray image generated by the color image and the surface normal vector generated by the depth image are used as multi-data mode information, and coordinate information features in the color image, the gray image and the surface normal vector are respectively extracted through a convolution-recurrent neural network to obtain a total feature of all coordinate information in the image as a reference feature.

As a still further scheme of the invention: in the step 2), image extraction is performed by using an improved background difference method of a Gaussian mixture model and an OHTA-based color segmentation extraction technology.

As a still further scheme of the invention: in the step 6), for the feature point set C, finding out N other positioning feature points closest to the feature point set C in the coordinate set B, and comparing the position change between the feature point set C and the positioning feature points.

As a still further scheme of the invention: in the step 5), the screenshot reserves 1-2 frames of data.

As a still further scheme of the invention: in the step 4), the coordinate calibration is performed on the whole picture from which the target object is separated, so as to form a background reference coordinate set B of the moving path.

Compared with the prior art, the invention has the beneficial effects that:

compared with the original identification technology, the target image and the environment image are independently marked, the whole image does not need to be subjected to whole comparison calculation, only the position change between the target image and a separated target point needs to be calculated, the memory is saved, the CPU resource is saved, and the identification efficiency is higher.

Detailed Description

The technical solution of the present invention will be described in further detail with reference to specific embodiments.

1) obtaining a picture with a fixed visual angle through monitoring equipment, obtaining an environment image, using a gray image generated by a color image and a surface normal vector generated by a depth image as multi-data mode information, respectively extracting coordinate information characteristics in the color image, the gray image and the surface normal vector through a convolution-recursion neural network, obtaining total characteristics of all coordinate information in the image as reference characteristics, and forming an index list of all coordinate information;

2) inputting a plurality of groups of contrast images, marking the approximate outline of the target object by utilizing an improved background difference method of a Gaussian mixture model and an OHTA-based color segmentation extraction technology, and separating the approximate outline from the whole picture;

4) carrying out coordinate calibration on the whole picture of the separated target object to form a background reference coordinate set B of the moving path;

5) the screenshot of the moved picture is reserved, 1-2 frame data of the screenshot is reserved, the pixel coordinate of the image to be recognized is normalized to obtain a pixel coordinate matrix of the image, and the feature point set C is compared with coordinate information in the image to obtain the position of the feature point set C in the picture;

6) and aiming at the characteristic point set C, finding out N other positioning characteristic points which are closest to the characteristic point set C in the coordinate set B, and obtaining the moving path of the characteristic point set C according to the position change between the characteristic point set C and the positioning characteristic points.

The working principle of the invention is as follows:

the picture is a fixed visual angle and a learning environment, 1-2 frame data are reserved, the data coming again is compared with the previous data, and the continuous moving object mark in the picture can be single or multiple, so that the method has the advantages that: compared with the prior art, the method saves the memory, saves the CPU resource and has higher identification efficiency.

While the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims

1. An intelligent efficient simultaneous recognition method for multiple objects is characterized by comprising the following steps:

4) forming a background reference coordinate set B of the moving path;

6) and obtaining the moving path of the feature point set C.

2. The method for efficiently and simultaneously identifying multiple intelligent objects according to claim 1, wherein in step 1), the gray image generated from the color image and the surface normal vector generated from the depth image are used together as multi-data mode information, and coordinate information features in the color image, the gray image and the surface normal vector are respectively extracted through a convolution-recursive neural network to obtain a total feature of all coordinate information in the image as a reference feature.

3. The method as claimed in claim 1, wherein in step 2), the image extraction is performed by using an improved background difference method of a mixture gaussian model and an OHTA-based color segmentation extraction technique.

4. The method as claimed in claim 1, wherein in step 6), for the feature point set C, N other positioning feature points closest to the feature point set C are found in the coordinate set B, and the position of the feature point set C is changed from the positioning feature points.

5. The method for efficiently and simultaneously recognizing multiple intelligent objects according to claim 1, wherein in the step 5), 1-2 frames of data are retained in the screenshot.

6. The method for efficiently and simultaneously identifying multiple intelligent objects according to claim 1, wherein in the step 4), the coordinates of the whole picture from which the target object is separated are calibrated to form a background reference coordinate set B of the moving path.