An intelligent camera for automatically capturing human faces and its recognition method

Abstract

The present invention relates to the field of data processing technology, and in particular to an intelligent camera for automatically capturing human faces and an identification method thereof. The method comprises collecting a first image, performing facial feature recognition on the first image, performing rectangular frame selection on the facial feature area, determining the facial area, obtaining a first corrected image, judging the first corrected image by a first threshold value to determine whether to name it a second image, judging the second image by a second threshold value to determine whether to name it a third image, and implementing corresponding processing methods based on the first corrected image, the second image, and the third image. The method achieves the effect of weighing multiple facial features, calculating capture priorities, and controlling the camera to automatically capture according to the capture priorities, thereby solving the problem of poor intelligent decision-making effect of automatic face capture on multiple facial features and reducing safety hazards in target scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to an intelligent camera for automatically capturing human faces and a recognition method thereof. Background Art

[0002] In today's society, security monitoring is becoming increasingly popular. In specific public or dangerous places, real-time monitoring of the environment and the flow of people is required. By identifying and comparing facial images, suspicious persons can be detected in a timely manner to prevent the occurrence of crimes and dangerous behaviors and improve the safety of the target places. However, in existing camera recognition methods, when multiple people appear in the monitoring screen, it is difficult to weigh the multiple facial features that appear and determine the face capture object. In particular, when the person is moving, it is difficult to control the camera to capture and track the object through the existing single face recognition method. As a result, the intelligent decision-making effect of automated face capture on multiple facial features is poor, and the level of hidden dangers is high. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems in the background technology and to propose an intelligent camera for automatically capturing human faces and a recognition method thereof.

[0004] In order to achieve the above object, the present invention adopts the following technical solution: a method for automatically capturing a face, comprising the following steps:

[0005] Capturing a video stream image and using it as a first image;

[0006] Performing facial feature recognition on the first image, and performing rectangular frame selection on the facial feature area to determine the facial area, thereby obtaining a first corrected image;

[0007] Obtaining the coordinate center point of the face area in the first corrected image and using it as the facial feature coordinate point, and counting the number of facial feature coordinate points in the first corrected image and using it as the total number of facial features;

[0008] Determine whether the total number of facial features in the first corrected image exceeds a first threshold, if the total number of facial features in the first corrected image is not greater than the first threshold, do not perform any operation; if the total number of facial features in the first corrected image is greater than the first threshold, name the first corrected image as the second image;

[0009] Analyze the second image to obtain an image analysis value, and determine whether the image analysis value exceeds a second threshold; if the image analysis value does not exceed the second threshold, do not perform any operation; if the image analysis value exceeds the second threshold, name the second image as a third image;

[0010] The method for analyzing the second image is:

[0011] Determine whether there are facial feature coordinate points in the monitoring quadrant; if there are facial feature coordinate points in the monitoring quadrant, mark the monitoring quadrant as the target quadrant; count the number of target quadrants and use it as the image analysis value;

[0012] performing preprocessing on the first corrected image, the second image, and the third image respectively, performing the first image processing on the second image, and performing the second image processing on the third image;

[0013] The methods for preprocessing the first corrected image, the second image, and the third image are respectively as follows:

[0014] Determine the total monitoring range based on the output format of the video stream, and divide the total monitoring range into several monitoring quadrants;

[0015] Obtain monitoring requirements and determine the effective monitoring scope;

[0016] Determining effective monitoring coordinate points constituting the effective monitoring range based on the effective monitoring range;

[0017] Determine whether the facial feature coordinate point belongs to the valid monitoring range. If the facial feature coordinate point belongs to the valid monitoring range, no operation is performed; otherwise, the monitoring quadrant where the facial feature coordinate point is located is marked as the target quadrant;

[0018] Mark the valid monitoring coordinate points in the target quadrant as target monitoring coordinate points;

[0019] The method of performing the first image processing on the second image is:

[0020] Determine whether there are facial feature coordinate points that do not fall within the effective monitoring range; if there are no facial feature coordinate points that do not fall within the effective monitoring range, capture the face corresponding to the facial feature coordinate points;

[0021] If there is a facial feature coordinate point that does not fall within the effective monitoring range, the facial feature coordinate point is marked as a pending feature coordinate point, the number of the pending feature coordinate points is counted to obtain the total number of the pending feature coordinate points; and data analysis is performed on the total number of the pending feature coordinate points;

[0022] If the total number of undetermined feature coordinate points is 1, the undetermined feature coordinate point is named the target feature coordinate point;

[0023] If the total number of undetermined feature coordinate points is not 1, calculate the distance between the target monitoring coordinate point and several undetermined feature coordinate points respectively, and use it as the control distance of the undetermined feature coordinate point. Compare the control distances of several undetermined feature coordinate points to obtain the maximum control distance, and name the undetermined feature coordinate point corresponding to the maximum control distance as the target feature coordinate point;

[0024] Control the camera according to the target monitoring coordinate points and target feature coordinate points;

[0025] The method of performing the second image processing on the third image is:

[0026] Mark facial feature coordinate points that are not within the effective monitoring range as key feature coordinate points, extract the facial information corresponding to the key coordinate points, put them into the face database for face information comparison, determine the identity information corresponding to the facial information, and determine the identity priority of the face based on the priority model;

[0027] Combining several key coordinate points in pairs to obtain several key feature coordinate point combinations;

[0028] Calculate the vectors of the two key feature coordinate points in the key feature coordinate point combination and use them as the key vector Vi (xi, yi) of the key feature coordinate point combination, where i is the number of the key feature coordinate point combination;

[0029] Randomly select two valid monitoring coordinate points that are not in the same monitoring quadrant to perform vector calculation and obtain the monitoring span vector Vc (xc, yc);

[0030] Determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) and obtain the vector judgment result;

[0031] The camera is controlled based on the identity priority and vector judgment results of the face.

[0032] As a further solution of the present invention, a method for regulating the camera according to the target monitoring coordinate points and the facial feature coordinate points is as follows:

[0033] By formula The corrected position vector V is calculated, where V0 is the basic position vector of the camera, CT is the coordinate parameter of the target feature coordinate point, and CM is the coordinate parameter of the target monitoring coordinate point;

[0034] The basic position vector is adjusted according to the modified position vector, and the coordinate parameters of the basic position vector, the target feature coordinate point, and the coordinate parameters of the target monitoring coordinate point are all updated according to the number of frames of the video stream.

[0035] As a further solution of the present invention, the identity information includes maintenance personnel, staff and general personnel;

[0036] The expression of the priority model is:

[0037] ; Y is the identity priority;

[0038] The method to determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) is:

[0039] Calculate the absolute value of the first component of Vi(xi,yi) respectively and the absolute value of the second component And, monitor the absolute value of the first component of the span vector Vc(xc,yc) and the absolute value of the second component ,judge No greater than and No greater than Are they established at the same time? No greater than and No greater than If both of them are true, then the judgment key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc); No greater than and No greater than If they are not true at the same time, it is determined that the judgment key vector Vi (xi, yi) does not satisfy the monitoring span vector Vc (xc, yc);

[0040] Outputting a coordinate determination value of a determination result based on a determination analysis model;

[0041] The expression of the decision analysis model is:

[0042] ; P is the coordinate judgment value of the two key feature coordinate points that constitute the key vector.

[0043] As a further solution of the present invention, a method for regulating camera movement according to the identity priority of the face and the monitoring span value is as follows:

[0044] Get several coordinate judgment values ​​of the same key feature coordinate point, and use the formula Calculate the capture priority Q of the key feature coordinate points;

[0045] Where k1 and k2 are the weight coefficients of identity priority Y and coordinate judgment value P respectively; j is the coordinate judgment number of the key feature coordinate point, j is a positive integer, j∈[1,n], and n is the total number of coordinate judgment values ​​of the key feature coordinate point;

[0046] Sort the capture priorities of the key feature coordinate points in descending order to obtain a priority sequence, and mark the key feature coordinate point at the first position in the priority sequence as the main reference coordinate point;

[0047] Obtain the monitoring quadrant of the primary reference coordinate point, mark the key feature coordinate point located in the monitoring quadrant as the secondary reference coordinate point, and mark the valid monitoring coordinate point located in the monitoring quadrant as the primary adjustment coordinate point, obtain the coordinate parameters of the primary adjustment coordinate point and record them as Cz;

[0048] Calculate the average coordinate point of the main reference coordinate point and several secondary reference coordinate points, and obtain the coordinate parameters of the average coordinate point , through the formula Calculate and obtain the corrected position vector V;

[0049] The basic position vector is regulated according to the modified position vector, and the basic position vector, the coordinate parameters of the main adjustment coordinate point and the coordinate parameters of the average coordinate point are all updated according to the number of frames of the video stream.

[0050] In addition, the present invention also discloses an intelligent camera for automatically capturing human faces, comprising an acquisition module, a recognition module, a feature determination module, a first analysis module, a second analysis module and an image processing module;

[0051] An acquisition module, configured to acquire an image from a video stream and use it as a first image;

[0052] A recognition module is used to perform facial feature recognition on the first image, and to perform rectangular selection on the facial feature area to determine the facial area and obtain a first corrected image;

[0053] a feature determination module, configured to obtain a coordinate center point of the face region in the first corrected image and use it as a facial feature coordinate point, and to count the number of facial feature coordinate points in the first corrected image and use it as the total number of facial features;

[0054] a first analysis module, configured to determine whether the total number of facial features in the first corrected image exceeds a first threshold, and if so, perform no operation; and if so, name the first corrected image as the second image;

[0055] a second analysis module configured to analyze the second image and determine whether a facial feature coordinate point exists in the monitoring quadrant; if a facial feature coordinate point exists in the monitoring quadrant, mark the monitoring quadrant as a target quadrant; count the number of target quadrants and use the result as an image analysis value to determine whether the image analysis value exceeds a second threshold; if the image analysis value does not exceed the second threshold, no operation is performed; if the image analysis value exceeds the second threshold, the second image is designated as a third image;

[0056] an image processing module, configured to pre-process the first corrected image, the second image, and the third image, perform the first image processing on the second image, and determine whether there are facial feature coordinate points that do not fall within the effective monitoring range; if no facial feature coordinate points do not fall within the effective monitoring range, then capture the face corresponding to the facial feature coordinate points;

[0057] If there is a facial feature coordinate point that does not fall within the effective monitoring range, the facial feature coordinate point is marked as a pending feature coordinate point, the number of the pending feature coordinate points is counted to obtain the total number of the pending feature coordinate points; and data analysis is performed on the total number of the pending feature coordinate points;

[0058] If the total number of undetermined feature coordinate points is 1, the undetermined feature coordinate point is named the target feature coordinate point;

[0059] If the total number of undetermined feature coordinate points is not 1, calculate the distance between the target monitoring coordinate point and several undetermined feature coordinate points respectively, and use it as the control distance of the undetermined feature coordinate point. Compare the control distances of several undetermined feature coordinate points to obtain the maximum control distance, and name the undetermined feature coordinate point corresponding to the maximum control distance as the target feature coordinate point;

[0060] Control the camera according to the target monitoring coordinate points and target feature coordinate points;

[0061] Performing the second image processing on the third image, marking facial feature coordinate points that are not within the effective monitoring range as key feature coordinate points, extracting facial information corresponding to the key coordinate points, placing the facial information into a face database for facial information comparison, determining identity information corresponding to the facial information, and determining the identity priority of the face based on the priority model;

[0062] Combining several key coordinate points in pairs to obtain several key feature coordinate point combinations;

[0063] Calculate the vectors of the two key feature coordinate points in the key feature coordinate point combination and use them as the key vector Vi (xi, yi) of the key feature coordinate point combination, where i is the number of the key feature coordinate point combination;

[0064] Randomly select two valid monitoring coordinate points that are not in the same monitoring quadrant to perform vector calculation and obtain the monitoring span vector Vc (xc, yc);

[0065] Determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) and obtain the vector judgment result;

[0066] The camera is controlled based on the identity priority and vector judgment results of the face.

[0067] Compared with the existing technology, the advantages of the present invention are: by collecting a first image, performing facial feature recognition on the first image, and performing a rectangular frame selection on the facial feature area, determining the facial area, and obtaining a first corrected image, the first corrected image is judged by a first threshold to decide whether to name it as a second image, the second image is judged by a second threshold to decide whether to name it as a third image, and corresponding processing methods are implemented according to the first corrected image, the second image and the third image, thereby achieving the effect of automatically capturing images with a single facial feature and locating facial feature coordinates, as well as weighing multiple facial features, calculating capture priority, and controlling the camera to automatically capture according to the capture priority, thereby solving the problem that the intelligent decision effect of automatic face capture on multiple facial features is poor, and reducing safety hazards in target occasions. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0070] Reference Figure 1 , a recognition method for automatically capturing a face, comprising:

[0071] Capturing a video stream image and using it as a first image;

[0072] Perform facial feature recognition on the first image, and perform rectangular selection on the facial feature area to determine the facial area, thereby obtaining a first corrected image. It should be noted that facial feature recognition can be implemented using existing facial recognition technologies such as neural networks, and the details will not be elaborated in detail.

[0073] Obtaining the coordinate center point of the face area in the first corrected image and using it as the facial feature coordinate point, and counting the number of facial feature coordinate points in the first corrected image and using it as the total number of facial features;

[0074] Determine whether the total number of facial features in the first corrected image exceeds a first threshold. If the total number of facial features in the first corrected image is not greater than the first threshold, no operation is performed. If the total number of facial features in the first corrected image is greater than the first threshold, the first corrected image is named the second image. The first threshold may be 1, that is, if the total number of facial features in the first corrected image exceeds 1, the first corrected image is named the second image.

[0075] Analyze the second image to obtain an image analysis value, and determine whether the image analysis value exceeds a second threshold; if the image analysis value does not exceed the second threshold, do not perform any operation; if the image analysis value exceeds the second threshold, name the second image as a third image;

[0076] The method for analyzing the second image is:

[0077] Determine whether there are facial feature coordinate points in the monitoring quadrant; if there are facial feature coordinate points in the monitoring quadrant, mark the monitoring quadrant as a target quadrant; count the number of target quadrants and use it as the image analysis value; it should be noted that the second threshold value can be 1, and if the number of target quadrants exceeds 1, the second image is named the third image;

[0078] Preprocessing the first corrected image, the second image, and the third image respectively;

[0079] The methods for preprocessing the first corrected image, the second image, and the third image are respectively as follows:

[0080] The total monitoring range is determined according to the output format of the video stream, and the total monitoring range is divided into a number of monitoring quadrants. For example, the monitoring quadrants can be divided into four equal areas of the total monitoring range, thereby obtaining four monitoring quadrants.

[0081] Obtain monitoring requirements and determine the effective monitoring range. It should be noted that the effective monitoring range may include, but is not limited to, areas where the camera can clearly identify human faces and capture behavioral characteristics of people.

[0082] Determine the effective monitoring coordinate points constituting the effective monitoring range based on the effective monitoring range. It should be noted that in mainstream surveillance videos, the output format is generally a rectangular 16:9 or 3:2 or similar video frame, and the effective monitoring coordinate points are the four coordinate points constituting the rectangular video frame.

[0083] Determine whether the facial feature coordinate point belongs to the valid monitoring range. If the facial feature coordinate point belongs to the valid monitoring range, no operation is performed; otherwise, the monitoring quadrant where the facial feature coordinate point is located is marked as the target quadrant;

[0084] Mark the valid monitoring coordinate points in the target quadrant as target monitoring coordinate points;

[0085] performing the first image processing on the second image;

[0086] The method of performing the first image processing on the second image is:

[0087] Determine whether there are facial feature coordinate points that do not fall within the effective monitoring range; if there are no facial feature coordinate points that do not fall within the effective monitoring range, capture the face corresponding to the facial feature coordinate points;

[0088] If there is a facial feature coordinate point that does not fall within the effective monitoring range, the facial feature coordinate point is marked as a pending feature coordinate point, the number of the pending feature coordinate points is counted to obtain the total number of the pending feature coordinate points; and data analysis is performed on the total number of the pending feature coordinate points;

[0089] If the total number of undetermined feature coordinate points is 1, the undetermined feature coordinate point is named the target feature coordinate point;

[0090] If the total number of undetermined feature coordinate points is not 1, calculate the distance between the target monitoring coordinate point and several undetermined feature coordinate points respectively, and use it as the control distance of the undetermined feature coordinate point. Compare the control distances of several undetermined feature coordinate points to obtain the maximum control distance, and name the undetermined feature coordinate point corresponding to the maximum control distance as the target feature coordinate point;

[0091] Control the camera according to the target monitoring coordinate points and target feature coordinate points;

[0092] The method for controlling the camera according to the target monitoring coordinate points and the facial feature coordinate points is as follows:

[0093] By formula The corrected position vector V is calculated, where V0 is the basic position vector of the camera, CT is the coordinate parameter of the target feature coordinate point, and CM is the coordinate parameter of the target monitoring coordinate point;

[0094] The basic position vector is regulated according to the modified position vector, and the coordinate parameters of the basic position vector, the target feature coordinate point, and the coordinate parameters of the target monitoring coordinate point are all updated according to the number of frames of the video stream;

[0095] performing the second image processing on the third image;

[0096] The method of performing the second image processing on the third image is:

[0097] Mark facial feature coordinate points that are not within the effective monitoring range as key feature coordinate points, extract the facial information corresponding to the key coordinate points, put them into the face database for face information comparison, determine the identity information corresponding to the facial information, and determine the identity priority of the face based on the priority model;

[0098] Combining a number of key coordinate points in pairs to obtain a number of key feature coordinate point combinations; in this embodiment, the pairwise combinations are all achieved using a permutation and combination method;

[0099] Calculate the vectors of the two key feature coordinate points in the key feature coordinate point combination and use them as the key vector Vi (xi, yi) of the key feature coordinate point combination, where i is the number of the key feature coordinate point combination;

[0100] Randomly select two valid monitoring coordinate points that are not in the same monitoring quadrant to perform vector calculation and obtain the monitoring span vector Vc (xc, yc);

[0101] Determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) and obtain the vector judgment result;

[0102] The camera is controlled based on the identity priority and vector judgment results of the face;

[0103] Identity information includes maintenance personnel, staff and general personnel;

[0104] The expression of the priority model is:

[0105] ; Y is the identity priority;

[0106] The method to determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) is:

[0107] Calculate the absolute value of the first component of Vi(xi,yi) respectively and the absolute value of the second component And, monitor the absolute value of the first component of the span vector Vc(xc,yc) and the absolute value of the second component ,judge No greater than and No greater than Are they established at the same time? No greater than and No greater than If both of them are true, then the judgment key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc); No greater than and No greater than If they are not true at the same time, it is determined that the judgment key vector Vi (xi, yi) does not satisfy the monitoring span vector Vc (xc, yc);

[0108] Outputting a coordinate determination value of a determination result based on a determination analysis model;

[0109] The expression of the decision analysis model is:

[0110] ; P is the coordinate determination value of the two key feature coordinate points that constitute the key vector; the method for controlling the camera movement according to the identity priority of the face and the monitoring span value is:

[0111] Get several coordinate judgment values ​​of the same key feature coordinate point, and use the formula Calculate the capture priority Q of the key feature coordinate points;

[0112] Where k1 and k2 are the weight coefficients of identity priority Y and coordinate judgment value P respectively; j is the coordinate judgment number of the key feature coordinate point, j is a positive integer, j∈[1,n], and n is the total number of coordinate judgment values ​​of the key feature coordinate point; it should be noted that the capture priority Q is related to the key vector formed by the key feature coordinate point and other key feature coordinate points, as well as the identity priority of the identity information of the corresponding face;

[0113] Sort the capture priorities of the key feature coordinate points in descending order to obtain a priority sequence, and mark the key feature coordinate point at the first position in the priority sequence as the main reference coordinate point;

[0114] Obtain the monitoring quadrant of the primary reference coordinate point, mark the key feature coordinate point located in the monitoring quadrant as the secondary reference coordinate point, and mark the valid monitoring coordinate point located in the monitoring quadrant as the primary adjustment coordinate point, obtain the coordinate parameters of the primary adjustment coordinate point and record them as Cz;

[0115] Calculate the average coordinate point of the main reference coordinate point and several secondary reference coordinate points, and obtain the coordinate parameters of the average coordinate point , through the formula Calculate and obtain the corrected position vector V;

[0116] The basic position vector is regulated according to the modified position vector, and the basic position vector, the coordinate parameters of the main adjustment coordinate point and the coordinate parameters of the average coordinate point are all updated according to the number of frames of the video stream.

[0117] Embodiment 2, an intelligent camera for automatically capturing human faces, comprising an acquisition module, a recognition module, a feature determination module, a first analysis module, a second analysis module and an image processing module;

[0118] An acquisition module, configured to acquire an image from a video stream and use it as a first image;

[0119] A recognition module is used to perform facial feature recognition on the first image, and to perform rectangular selection on the facial feature area to determine the facial area and obtain a first corrected image;

[0120] a feature determination module, configured to obtain a coordinate center point of the face region in the first corrected image and use it as a facial feature coordinate point, and to count the number of facial feature coordinate points in the first corrected image and use it as the total number of facial features;

[0121] a first analysis module, configured to determine whether the total number of facial features in the first corrected image exceeds a first threshold, and if so, perform no operation; and if so, name the first corrected image as the second image;

[0122] a second analysis module, configured to analyze the second image, obtain an image analysis value, and determine whether the image analysis value exceeds a second threshold; if the image analysis value does not exceed the second threshold, no operation is performed; if the image analysis value exceeds the second threshold, the second image is named a third image;

[0123] an image processing module, configured to pre-process the first corrected image, the second image, and the third image, perform the first image processing on the second image, and perform the second image processing on the third image;

[0124] In this embodiment, the smart camera adopts a modern surveillance camera with functions such as steering and rotation.

[0125] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for automatically capturing a face, characterized by: This includes the following methods: Capturing a video stream image and using it as a first image; Performing facial feature recognition on the first image, and performing rectangular frame selection on the facial feature area to determine the facial area, thereby obtaining a first corrected image; Obtaining the coordinate center point of the face area in the first corrected image and using it as the facial feature coordinate point, and counting the number of facial feature coordinate points in the first corrected image and using it as the total number of facial features; Determine whether the total number of facial features in the first corrected image exceeds a first threshold, if the total number of facial features in the first corrected image is not greater than the first threshold, do not perform any operation; if the total number of facial features in the first corrected image is greater than the first threshold, name the first corrected image as the second image; Analyze the second image to obtain an image analysis value, and determine whether the image analysis value exceeds a second threshold; If the image analysis value does not exceed the second threshold, no operation is performed; if the image analysis value exceeds the second threshold, the second image is named the third image; The method for analyzing the second image is: Determine whether there are facial feature coordinate points in the monitoring quadrant; if there are facial feature coordinate points in the monitoring quadrant, mark the monitoring quadrant as the first target quadrant; count the number of first target quadrants and use it as the image analysis value; performing preprocessing on the first corrected image, the second image, and the third image respectively, performing the first image processing on the second image, and performing the second image processing on the third image; The methods for preprocessing the first corrected image, the second image, and the third image are respectively as follows: Determine the total monitoring range based on the output format of the video stream, and divide the total monitoring range into several monitoring quadrants; Obtain monitoring requirements and determine the effective monitoring scope; Determining effective monitoring coordinate points constituting the effective monitoring range based on the effective monitoring range; Determine whether the facial feature coordinate point belongs to the valid monitoring range. If the facial feature coordinate point belongs to the valid monitoring range, no operation is performed; otherwise, the monitoring quadrant where the facial feature coordinate point is located is marked as the second target quadrant; Mark the valid monitoring coordinate point in the second target quadrant as the target monitoring coordinate point; The method of performing the first image processing on the second image is: Determine whether there are facial feature coordinate points that do not fall within the effective monitoring range; if there are no facial feature coordinate points that do not fall within the effective monitoring range, capture the face corresponding to the facial feature coordinate points; If there is a facial feature coordinate point that does not belong to the effective monitoring range, the facial feature coordinate point is marked as a pending feature coordinate point, and the number of pending feature coordinate points is counted to obtain the total number of pending feature coordinate points; Perform data analysis on the total number of feature coordinate points to be determined; If the total number of undetermined feature coordinate points is 1, the undetermined feature coordinate point is named the target feature coordinate point; If the total number of undetermined feature coordinate points is not 1, calculate the distance between the target monitoring coordinate point and several undetermined feature coordinate points respectively, and use it as the control distance of the undetermined feature coordinate point. Compare the control distances of several undetermined feature coordinate points to obtain the maximum control distance, and name the undetermined feature coordinate point corresponding to the maximum control distance as the target feature coordinate point; Control the camera according to the target monitoring coordinate points and target feature coordinate points; The method of performing the second image processing on the third image is: Mark facial feature coordinate points that are not within the effective monitoring range as key feature coordinate points, extract the facial information corresponding to the key feature coordinate points, put them into the face database for face information comparison, determine the identity information corresponding to the facial information, and determine the identity priority of the face based on the priority model; Combining a number of key feature coordinate points in pairs to obtain a number of key feature coordinate point combinations; Calculate the vectors of the two key feature coordinate points in the key feature coordinate point combination and use them as the key vector Vi (xi, yi) of the key feature coordinate point combination, where i is the number of the key feature coordinate point combination; Randomly select two valid monitoring coordinate points that are not in the same monitoring quadrant to perform vector calculation and obtain the monitoring span vector Vc (xc, yc); Determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) and obtain the vector judgment result; The camera is controlled based on the identity priority and vector judgment results of the face.

2. The method for automatically capturing a face according to claim 1, wherein: Identity information includes maintenance personnel, staff and general personnel; The expression of the priority model is: ; Y is the identity priority; The method to determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) is: Calculate the absolute value of the first component of Vi(xi,yi) respectively and the absolute value of the second component And, monitor the absolute value of the first component of the span vector Vc(xc,yc) and the absolute value of the second component ,judge No greater than and No greater than Are they established at the same time? No greater than and No greater than If both of them are true, then the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc); No greater than and No greater than If they are not true at the same time, it is determined that the judgment key vector Vi (xi, yi) does not satisfy the monitoring span vector Vc (xc, yc); Output the vector judgment result of the key vector Vi (xi, yi) based on the judgment analysis model; The expression of the decision analysis model is: ; P is the coordinate judgment value of the two key feature coordinate points that constitute the key vector, which is used as the vector judgment result of the key vector Vi (xi, yi).

3. An intelligent camera for automatically capturing human faces, applying the automatic face recognition method according to claim 2, characterized in that: It includes an acquisition module, an identification module, a feature determination module, a first analysis module, a second analysis module and an image processing module; An acquisition module, configured to acquire an image from a video stream and use it as a first image; A recognition module is used to perform facial feature recognition on the first image, and to perform rectangular selection on the facial feature area to determine the facial area and obtain a first corrected image; a feature determination module, configured to obtain a coordinate center point of the face region in the first corrected image and use it as a facial feature coordinate point, and to count the number of facial feature coordinate points in the first corrected image and use it as the total number of facial features; a first analysis module, configured to determine whether the total number of facial features in the first corrected image exceeds a first threshold, and if so, perform no operation; and if so, name the first corrected image as the second image; A second analysis module is used to analyze the second image to determine whether there are facial feature coordinate points in the monitoring quadrant; If there is a facial feature coordinate point in the monitoring quadrant, the monitoring quadrant is marked as the first target quadrant; Counting the number of the first target quadrant and using it as an image analysis value, and determining whether the image analysis value exceeds a second threshold; If the image analysis value does not exceed the second threshold, no operation is performed; if the image analysis value exceeds the second threshold, the second image is named the third image; The image processing module is used to pre-process the first corrected image, the second image, and the third image respectively. The method of pre-processing the first corrected image, the second image, and the third image respectively is: Determine the total monitoring range based on the output format of the video stream, and divide the total monitoring range into several monitoring quadrants; Obtain monitoring requirements and determine the effective monitoring scope; Determining effective monitoring coordinate points constituting the effective monitoring range based on the effective monitoring range; Determine whether the facial feature coordinate point belongs to the valid monitoring range. If the facial feature coordinate point belongs to the valid monitoring range, no operation is performed; otherwise, the monitoring quadrant where the facial feature coordinate point is located is marked as the second target quadrant; Mark the valid monitoring coordinate point in the second target quadrant as the target monitoring coordinate point; Performing the first image processing on the second image to determine whether there are facial feature coordinate points that do not fall within the effective monitoring range; if there are no facial feature coordinate points that do not fall within the effective monitoring range, capturing the face corresponding to the facial feature coordinate points; If there is a facial feature coordinate point that does not belong to the effective monitoring range, the facial feature coordinate point is marked as a pending feature coordinate point, and the number of pending feature coordinate points is counted to obtain the total number of pending feature coordinate points; Perform data analysis on the total number of feature coordinate points to be determined; If the total number of undetermined feature coordinate points is 1, the undetermined feature coordinate point is named the target feature coordinate point; If the total number of undetermined feature coordinate points is not 1, calculate the distance between the target monitoring coordinate point and several undetermined feature coordinate points respectively, and use it as the control distance of the undetermined feature coordinate point. Compare the control distances of several undetermined feature coordinate points to obtain the maximum control distance, and name the undetermined feature coordinate point corresponding to the maximum control distance as the target feature coordinate point; Control the camera according to the target monitoring coordinate points and target feature coordinate points; Performing the second image processing on the third image, marking facial feature coordinate points that are not within the effective monitoring range as key feature coordinate points, extracting facial information corresponding to the key feature coordinate points, placing the facial information into a face database for facial information comparison, determining identity information corresponding to the facial information, and determining the identity priority of the face based on the priority model; Combining a number of key feature coordinate points in pairs to obtain a number of key feature coordinate point combinations; Calculate the vectors of the two key feature coordinate points in the key feature coordinate point combination and use them as the key vector Vi (xi, yi) of the key feature coordinate point combination, where i is the number of the key feature coordinate point combination; Randomly select two valid monitoring coordinate points that are not in the same monitoring quadrant to perform vector calculation and obtain the monitoring span vector Vc (xc, yc); Determine whether the key vector Vi (xi, yi) satisfies the monitoring span vector Vc (xc, yc) and obtain the vector judgment result; The camera is controlled based on the identity priority and vector judgment results of the face.

Images