阅读说明：本技术 一种摄像头人脸识别聚合优化方法 (Camera face recognition aggregation optimization method ) 是由 徐庆庆 张彪 于 2021-09-01 设计创作，主要内容包括：本发明公开了一种摄像头人脸识别聚合优化方法,包括以下步骤；步骤一、摄像头将视频码流上传到服务端,服务端对视频码流进行抽帧,将抽帧得到的图片进行人脸监测,通过检测分值得出top3图片；步骤二、服务端将top3中的每个图片与已有的人脸库中的人脸逐一比较,通过特有的人脸算法打出对应一组分值；步骤三、提取最高分值与其对应的人脸库中的人脸,平台配置两个不同的阈值,用于控制该人脸图片的处理结果；步骤四、将人脸图片分值与配置好的阈值进行比较,高于阈值1视为归类到对应人脸,低于阈值2视为注册成新人脸,两者中间视为丢弃。本发明识别精准度更高,同一个人更少得被识别成多个人可以主动对识别结果进行修正,使得后续识别更准确。(The invention discloses a camera face identification aggregation optimization method, which comprises the following steps of; the method comprises the steps that firstly, a camera uploads a video code stream to a server, the server performs frame extraction on the video code stream, a picture obtained through frame extraction is subjected to face monitoring, and a top3 picture is obtained through detection scores; step two, the server compares each picture in top3 with the face in the existing face library one by one, and a group of corresponding scores are printed through a specific face algorithm; extracting the face in the face library corresponding to the highest score, and configuring two different thresholds for controlling the processing result of the face picture by the platform; and step four, comparing the face picture score with a configured threshold, wherein a face picture score higher than the threshold 1 is regarded as a classified corresponding face, a face picture score lower than the threshold 2 is regarded as a registered new face, and the face picture score is regarded as discarded between the two faces. The method has higher identification precision, and the same person can be identified into a plurality of persons less and can actively correct the identification result, so that the subsequent identification is more accurate.)

1. A camera face recognition aggregation optimization method is characterized by comprising the following steps: comprises the following steps;

the method comprises the steps that firstly, a camera uploads a video code stream to a server, the server performs frame extraction on the video code stream, a picture obtained through frame extraction is subjected to face monitoring, and a top3 picture is obtained through detection scores;

step two, the server compares each picture in top3 with the face in the existing face library one by one, and a group of corresponding scores are printed through a specific face algorithm;

extracting the face in the face library corresponding to the highest score, and configuring two different thresholds for controlling the processing result of the face picture by the platform;

step four, comparing the face picture score with a configured threshold, wherein a face picture score higher than the threshold 1 is regarded as a classified corresponding face, a face picture score lower than the threshold 2 is regarded as a registered new face, and the face picture score and the threshold are discarded;

step five, if the picture is classified into the existing face, updating the corresponding face feature library; if the picture needs to be registered as a new face, a new face is added in the face library and a feature is added in the face feature library;

step six, through the client, the user can manually aggregate different faces, the aggregated combined face feature library is subjected to combining processing, and the combined faces are deleted;

and step seven, simultaneously, the server side creates the deleted face library id and the index of the new merged face library, and the user can be guided to the merged face library to access the deleted face so as to obtain correct information.

2. The camera face recognition aggregation optimization method according to claim 1, wherein: in the first step, when the camera carries out face recognition, a preset position of the camera is firstly obtained, the camera is restored to the position when the recognition is started each time, then a space coordinate where the face is located is obtained from a server, then a relative relation between the position where the face is located and the initial position of the camera is calculated through the space coordinate, and the motion of the camera is controlled, so that the face is located in the center of the visual field of the camera.

3. The camera face recognition aggregation optimization method according to claim 2, wherein: and during the detection of the acquired face, if the face is detected, the face is identified, otherwise, the detection is continued.

4. The camera face recognition aggregation optimization method according to claim 1, wherein: in the fourth step, a lowest threshold and a highest threshold of the confidence of the face may be preset, and a minimum threshold and a maximum threshold of the angle value of the face may be preset.

Technical Field

The invention relates to the technical field of face recognition, in particular to a camera face recognition aggregation optimization method.

Background

Along with the continuous popularization of security cameras and the continuous improvement of the intelligent degree of the security cameras, more and more face recognition is applied to products of the type, however, the security cameras are influenced by some factors such as light rays and angles, the problems of mistaken recognition, low recognition accuracy and the like often occur in the cameras, the problem that the mistaken recognition cannot be well handled by the existing scheme on the market is solved, an optimization scheme is needed to ensure the accuracy and the use experience of the cameras for face recognition, the scheme discloses a method for carrying out aggregation management on recognized face information in the face recognition, and the problem of the mistaken recognition of the face in the security cameras is solved.

Disclosure of Invention

The invention aims to provide a camera face recognition aggregation optimization method to solve the problems in the background technology.

In order to achieve the purpose, the invention provides the following technical scheme: a camera face recognition aggregation optimization method comprises the following steps;

the method comprises the steps that firstly, a camera uploads a video code stream to a server, the server performs frame extraction on the video code stream, a picture obtained through frame extraction is subjected to face monitoring, and a top3 picture is obtained through detection scores;

step two, the server compares each picture in top3 with the face in the existing face library one by one, and a group of corresponding scores are printed through a specific face algorithm;

extracting the face in the face library corresponding to the highest score, and configuring two different thresholds for controlling the processing result of the face picture by the platform;

step four, comparing the face picture score with a configured threshold, wherein a face picture score higher than the threshold 1 is regarded as a classified corresponding face, a face picture score lower than the threshold 2 is regarded as a registered new face, and the face picture score and the threshold are discarded;

step five, if the picture is classified into the existing face, updating the corresponding face feature library; if the picture needs to be registered as a new face, a new face is added in the face library and a feature is added in the face feature library;

step six, through the client, the user can manually aggregate different faces, the aggregated combined face feature library is subjected to combining processing, and the combined faces are deleted;

and step seven, simultaneously, the server side creates the deleted face library id and the index of the new merged face library, and the user can be guided to the merged face library to access the deleted face so as to obtain correct information.

Preferably, in the first step, when the camera performs face recognition, a preset position of the camera is obtained first, the camera is restored to the position when the recognition starts each time, then a spatial coordinate of the face is obtained from the server, then a relative relationship between the position of the face and the initial position of the camera is calculated through the spatial coordinate, and the movement of the camera is controlled, so that the face is located in the center of the field of view of the camera.

Preferably, when the acquired face is detected, if the face is detected, the face is identified, otherwise, the detection is continued.

Preferably, in the fourth step, a lowest threshold and a highest threshold of the confidence of the face may be preset, and a minimum threshold and a maximum threshold of the angle value of the face may be preset.

The invention provides a camera face identification aggregation optimization method, which has the beneficial effects that: the method has higher identification precision, and the same person can be identified into a plurality of persons less and can actively correct the identification result, so that the subsequent identification is more accurate.

Drawings

FIG. 1 is a schematic flow diagram of the present invention;

fig. 2 is a flowchart of the manual face aggregation process of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Referring to fig. 1-2, the present invention provides a technical solution: a camera face recognition aggregation optimization method comprises the following steps;

the method comprises the steps that firstly, a camera uploads a video code stream to a server, the server performs frame extraction on the video code stream, a picture obtained through frame extraction is subjected to face monitoring, and a top3 picture is obtained through detection scores;

when the camera carries out face recognition, firstly obtaining a preset position of the camera, restoring the camera to the position when the recognition is started each time, then obtaining a space coordinate where the face is located from a server, then calculating a relative relation between the position where the face is located and an initial position of the camera according to the space coordinate, controlling the motion of the camera to enable the face to be located at the center of a visual field of the camera, and when the obtained face is detected, if the face is detected, starting the recognition, otherwise, continuing the detection;

step two, the server compares each picture in top3 with the face in the existing face library one by one, and a group of corresponding scores are printed through a specific face algorithm;

extracting the face in the face library corresponding to the highest score, and configuring two different thresholds for controlling the processing result of the face picture by the platform;

step four, comparing the face picture score with a configured threshold, wherein a face picture score higher than the threshold 1 is regarded as a classified corresponding face, a face picture score lower than the threshold 2 is regarded as a registered new face, and the face picture score and the threshold are discarded;

the minimum threshold and the maximum threshold of the confidence coefficient of the face can be preset, and the minimum threshold and the maximum threshold of the angle value of the face can be preset;

step five, if the picture is classified into the existing face, updating the corresponding face feature library; if the picture needs to be registered as a new face, a new face is added in the face library and a feature is added in the face feature library;

step six, through the client, the user can manually aggregate different faces, the aggregated combined face feature library is subjected to combining processing, and the combined faces are deleted;

and step seven, simultaneously, the server side creates the deleted face library id and the index of the new merged face library, and the user can be guided to the merged face library to access the deleted face so as to obtain correct information.

In the embodiment, the method is divided into two parts, namely automatic aggregation and active aggregation, in an actual scene, due to factors such as ambient light and human face angles, a collected human face picture is fuzzy or incomplete in face, the human face picture is easy to be recognized by mistake or discarded by Wu, in order to reduce the occurrence of the situation, after a specific human face algorithm is adopted, the captured picture needs to be aggregated, and different operations such as classification, registration as a new person and discarding are finally executed, wherein 2 different human face scoring thresholds are involved and are used for controlling the accuracy of the final operation;

(1) the platform performs frame extraction on the code stream uploaded by the camera, in order to control the recognition efficiency and accuracy, a GOP is adopted to extract 6 frames in the scheme, and then the 6 pictures are subjected to face detection to obtain top 3;

(2) comparing each picture in the screened top3 with the face of the existing face library through a server face algorithm, and marking a group of scores;

(3) determining the highest score, selecting a comparison face corresponding to the highest score, and comparing the score with two thresholds configured by the platform, wherein the threshold 1 is greater than the threshold 2, the face classified into the corresponding face is considered to be higher than the threshold 1, the face registered as a new face is considered to be lower than the threshold 2, and the middle of the two faces is discarded;

(4) the method comprises the steps that the maximum 9 face features are allowed under each face, each feature corresponds to a score, if the face is finally considered to be classified, the picture is added into a feature library corresponding to the face, if the number of the features in the feature library exceeds 9 features, the feature with the lowest score is removed, and the process is automatic aggregation;

(5) for the recognized face, if the situation that a person is mistakenly recognized into a plurality of persons exists, the user can manually aggregate different faces into one person through the client, a corresponding face feature library can be expanded, the maximum number of features is 50, the recognition can be performed with the features in the feature library every time, the more the features are, and the more the face recognition is.

(6) The accuracy of face recognition can be improved to a great extent through processes of threshold control, feature library expansion, automatic aggregation, manual aggregation and the like;

(7) and deleting the original merged face after the face merging, and creating an index of a deleted face id and a merged face library to ensure that the deleted face can correctly point to the merged face, so that the deleted face can still correctly acquire related information in front-end display.

Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.