Monitoring data screening method, system, readable storage medium and computer device

Abstract

The present invention proposes a monitoring data screening method, system, readable storage medium, and computer device. The method includes: obtaining identity information uploaded by a person to be supervised, and retrieving a facial image corresponding to the person to be supervised from a preset facial image database based on the identity information; identifying facial features contained in the monitoring image based on a final facial recognition model; determining whether the facial features contained in the facial image corresponding to the person to be supervised are consistent with the facial features contained in the monitoring image; if they are consistent, determining that the person to be supervised has passed through a target area, and summarizing the names of all target areas passed by the person to be supervised, so as to generate a behavior monitoring video corresponding to the person to be supervised based on all the names of the target areas passed by the person to be supervised. The monitoring data screening method proposed by the present invention can automatically screen out monitoring videos about the person to be supervised, greatly reducing labor intensity, and effectively avoiding information loss.

Description

Technical Field

[0001] The present invention relates to the field of monitoring management technology, and in particular to a monitoring data screening method, system, readable storage medium and computer equipment. Background Art

[0002] With the development of science and technology, monitoring systems play an important role in people's daily lives, greatly improving the safety of the living environment.

[0003] In the case handling area of ​​the procuratorate, judicial police are generally required to conduct real-time personnel monitoring of criminal suspects entering the case handling area. At the same time, multiple surveillance cameras will be installed in the inspector's case handling area to conduct full-time video surveillance of criminal suspects.

[0004] When a criminal suspect enters the case handling area, bailiffs need to keep an eye on the suspect's movements to prevent him from engaging in extreme behavior. However, long periods of supervision can easily fatigue the bailiffs, leading to inadequate supervision and increasing the probability of the suspect engaging in extreme behavior. To prevent unnecessary disputes, all surveillance videos need to be manually screened to restore the scene where the extreme behavior occurred. However, the manual screening and screening of massive videos results in high labor intensity and is prone to information loss. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to propose a monitoring data screening method, system, readable storage medium and computer equipment to solve the problems of high labor intensity and easy loss of information caused by the traditional evidence collection method of manually screening and checking massive videos.

[0006] A monitoring data screening method proposed in the present invention is applied to a monitoring management platform, and the method includes:

[0007] When the person to be supervised enters the supervision space, the identity information uploaded by the person to be supervised is obtained, and a facial image corresponding to the person to be supervised is retrieved from a preset facial image database according to the identity information;

[0008] Inputting the surveillance images of the target area into the trained final face recognition model in sequence, so as to identify facial features contained in the surveillance images according to the final face recognition model, wherein the facial features include facial contour features and multiple facial features, and the supervision space includes multiple target areas;

[0009] Determining whether the facial features of the facial image corresponding to the person to be supervised are consistent with the facial features of the monitoring image;

[0010] If the facial features contained in the facial image are consistent with the facial features contained in the surveillance image, it is determined that the person to be supervised has passed through the target area, and the names of all the target areas passed by the person to be supervised are summarized to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all the target areas passed by the person to be supervised.

[0011] In summary, according to the above-mentioned monitoring data screening method, by automatically identifying all monitoring images to confirm whether the monitoring image is related to the person to be supervised, a monitoring video of the person to be supervised is accurately compiled, eliminating the need for manual screening and investigation of massive amounts of videos, which greatly reduces labor intensity. Specifically, the identity information of the person to be supervised is first obtained, and the facial image of the person to be supervised is retrieved from a preset image library based on the identity information. Then, all monitoring images of the target area are sequentially input into the trained recognition model to obtain multiple facial features contained in the monitoring image. The multiple facial features in the monitoring image are then compared with the multiple facial features in the facial images of the person to be supervised previously retrieved to determine whether the person in the monitoring image is the person to be supervised. If so, it indicates that the person to be supervised has passed through the target area. The names of all target areas that the person to be supervised has passed through in the target space are then obtained, and a monitoring video of the person to be supervised's behavior is generated, solving the problems of high manual labor intensity and easy information loss caused by manual data screening and investigation.

[0012] Furthermore, the step of constructing the final face recognition model includes:

[0013] Acquire multiple copies of historical image information, perform face detection on the historical image information to obtain face position information, and annotate the historical image information according to the face position information;

[0014] The historical image information is cropped according to the annotation results to obtain a target image containing facial information of the person to be supervised, and multiple facial key points in the target image are extracted, wherein the facial key points include the center of the binocular eyes and the center of the lips.

[0015] Furthermore, after the step of cropping the historical image information according to the annotation results to obtain a target image containing the facial information of the person to be supervised, and extracting a plurality of facial key points in the target image, the method further includes:

[0016] Creating a canvas and marking a plurality of target positions on the canvas, wherein the target positions correspond to the facial key points one by one, so as to map the pixel information of each facial key point to the corresponding target position one by one according to the position information of the binocular center and the position information of the lip center, thereby obtaining a new canvas on which the facial key points are mapped;

[0017] Obtaining the position information and pixel information of each facial feature in the target image, so as to calculate the mapping position information of each facial feature in the new canvas based on the position information of the facial feature in the target image;

[0018] According to the mapping position information and pixel information of each facial feature in the target image, the facial features are mapped one by one to the new canvas to obtain a face-aligned image.

[0019] Furthermore, after the step of mapping any point in the target image to the new canvas according to the mapping position information and pixel information of the point to obtain the face aligned image, the step further includes:

[0020] The pixel values ​​of all target points in the face alignment image are obtained to perform illumination compensation on each target point according to the pixel value of each target point, and the face alignment image after illumination compensation is input into the initial face recognition model for training to obtain the final face recognition model.

[0021] Furthermore, the step of determining whether the facial features contained in the facial image corresponding to the person to be supervised are consistent with the facial features contained in the monitoring image includes:

[0022] Obtaining facial feature distance values ​​and facial contour feature distance values ​​of the face image and the surveillance image;

[0023] Determine whether the facial feature distance value and the facial contour feature distance value are both less than a preset feature distance threshold.

[0024] Furthermore, if the facial features included in the facial image are consistent with the facial features included in the surveillance image, then determining that the person to be supervised has passed through the target area, and summarizing the names of all target areas passed by the person to be supervised to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all target areas passed by the person to be supervised includes the following steps:

[0025] If it is determined that the facial feature distance value and the facial contour feature distance value are both less than the preset feature distance threshold, then it is determined that the surveillance image recognized in the current frame is consistent with the facial image of the person to be supervised, and the name of the target area corresponding to the surveillance image is obtained within a first preset time;

[0026] The surveillance images are numbered according to the target area name, and the next frame of surveillance images is extracted from the surveillance video of the target space for repeated recognition;

[0027] All surveillance images containing numbers are retrieved, and the retrieved surveillance images are sorted and summarized according to the numbering results to obtain the target video.

[0028] Furthermore, the method further comprises:

[0029] The pixel values ​​of all target points are compensated according to the following formula:

[0030] X pt =(X pi -127.5) / 128

[0031] Among them, X pt Indicates the pixel value of the target point after illumination compensation, X pi Indicates the pixel value of the target point before illumination compensation.

[0032] A monitoring data screening system according to an embodiment of the present invention is applied to a monitoring management platform, and the system includes:

[0033] An identity information acquisition module is used to acquire the identity information uploaded by the person to be supervised, and to retrieve a facial image corresponding to the person to be supervised from a preset facial image database based on the identity information;

[0034] A facial feature recognition module is configured to sequentially input surveillance images of a target area into a trained final facial recognition model to identify facial features contained in the surveillance images based on the final facial recognition model. The facial features include facial contour features and multiple facial features. The supervision space includes multiple target areas.

[0035] A face detection module is used to determine whether the facial features of the face image corresponding to the supervised person are consistent with the facial features of the monitoring image;

[0036] The behavior monitoring video generation module is used to determine that the person to be supervised has passed through the target area if the facial features contained in the facial image are consistent with the facial features contained in the monitoring image, and to summarize the names of all target areas passed by the person to be supervised, so as to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all target areas passed by the person to be supervised.

[0037] Another aspect of the present invention provides a readable storage medium, comprising one or more programs stored in the readable storage medium, which implement the monitoring data screening method as described above when the programs are executed.

[0038] Another aspect of the present invention provides a computer device, comprising a memory and a processor, wherein:

[0039] The memory is used to store computer programs;

[0040] When the processor is used to execute the computer program stored in the memory, the monitoring data screening method as described above is implemented.

[0041] Additional aspects and advantages of the present invention will be set forth in part in the following description and, in part, will be obvious from the following description, or may be learned through embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the monitoring data screening method proposed in the first embodiment of the present invention;

[0043] Figure 2 This is a flow chart of a monitoring data screening method proposed in a second embodiment of the present invention;

[0044] Figure 3 This is a flowchart of constructing the final face recognition model in the second embodiment of the present invention;

[0045] Figure 4 This is a schematic structural diagram of a monitoring data screening system proposed in the third embodiment of the present invention.

[0046] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0047] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0049] See also Figure 1 , which is a flow chart of a monitoring data screening method in a first embodiment of the present invention, is applied to a monitoring management platform. The method includes steps S01 to S04, wherein:

[0050] Step S01: obtaining identity information uploaded by the person to be supervised, and retrieving a facial image corresponding to the person to be supervised from a preset facial image database according to the identity information;

[0051] It is understandable that before starting data screening, all persons to be supervised will establish a relevant facial image database in advance. The preset facial image database includes the identity information of multiple persons to be supervised and the facial images corresponding to each identity information. The preset facial image database is then pre-stored in the monitoring management platform. When a person to be supervised enters the supervision space, the supervision space is the case handling area, interrogation room, toilet and other places. An identity information acquisition module will be provided at the door of the supervision space, so that the person to be supervised will first need to upload his or her identity information when preparing to enter the supervision space. In this embodiment, the identity information can be fingerprint information, ID card information and other unique identification information.

[0052] Step S02: sequentially inputting the surveillance images of the target area into the trained final face recognition model to identify facial features contained in the surveillance images according to the final face recognition model, wherein the facial features include facial contour features and multiple facial features, and the supervision space includes multiple target areas;

[0053] It should be noted that the monitoring space includes multiple target areas, and each target area generally has at least one camera installed for shooting video. However, since the video shot by each camera contains many images, and the video shot by the same camera does not know which image contents correspond to the supervised persons, based on this, the monitoring management platform will identify the video shot by each camera frame by frame, that is, input the monitoring image of each target area into the constructed final face recognition model to analyze the facial features contained in the monitoring image, where the monitoring image is extracted frame by frame from the video shot by the camera.

[0054] Step S03: determining whether the facial features of the facial image corresponding to the person to be supervised are consistent with the facial features of the monitoring image;

[0055] In order to determine whether the identity of the person in the currently analyzed surveillance image is the person to be supervised, it is necessary to compare the facial features analyzed from the surveillance image with the facial features in the previously retrieved facial image.

[0056] It should be noted that the facial features of the facial image also need to be recognized and obtained by the final face recognition model.

[0057] Step S04: If the facial features contained in the facial image are consistent with the facial features contained in the surveillance image, it is determined that the person to be supervised has passed through the target area, and the names of all the target areas passed by the person to be supervised are summarized to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all the target areas passed by the person to be supervised.

[0058] It should be noted that if the facial feature comparison is inconsistent, it means that the surveillance image is not of the person to be supervised. At this time, the next frame of surveillance image will be extracted for repeated analysis. If the facial feature comparison is consistent, it means that the surveillance image is of the person to be supervised. At this time, the monitoring management platform will determine that the person to be supervised has passed through this target area. This will continue until all frames of surveillance images are analyzed, and the names of all target areas passed by the person to be supervised can be accurately known. The surveillance images that have passed through these target area names will then be summarized to obtain the surveillance video of the person to be supervised, thereby replacing the manual screening and investigation of massive surveillance data one by one.

[0059] In summary, according to the above-mentioned monitoring data screening method, by automatically identifying all monitoring images to confirm whether the monitoring image is related to the person to be supervised, a monitoring video of the person to be supervised is accurately compiled, eliminating the need for manual screening and investigation of massive amounts of videos, which greatly reduces labor intensity. Specifically, the identity information of the person to be supervised is first obtained, and the facial image of the person to be supervised is retrieved from a preset image library based on the identity information. Then, all monitoring images of the target area are sequentially input into the trained recognition model to obtain multiple facial features contained in the monitoring image. The multiple facial features in the monitoring image are then compared with the multiple facial features in the facial images of the person to be supervised previously retrieved to determine whether the person in the monitoring image is the person to be supervised. If so, it indicates that the person to be supervised has passed through the target area. The names of all target areas that the person to be supervised has passed through in the target space are then obtained, and a monitoring video of the person to be supervised's behavior is generated, solving the problems of high manual labor intensity and easy information loss caused by manual data screening and investigation.

[0060] See also Figure 2 , which is a flow chart of a monitoring data screening method in a second embodiment of the present invention, is applied to a monitoring management platform. The method includes steps S101 to S107, wherein:

[0061] Step S101: obtaining identity information uploaded by the person to be supervised, and retrieving a facial image corresponding to the person to be supervised from a preset facial image database according to the identity information;

[0062] Step S102: sequentially inputting the surveillance images of the target area into the trained final face recognition model to identify facial features contained in the surveillance images according to the final face recognition model, wherein the facial features include facial contour features and multiple facial features, and the supervision space includes multiple target areas;

[0063] Please note that Figure 3The process of constructing the final face recognition model includes steps S1021 to S1022, wherein:

[0064] Step S1021: Acquire multiple historical image information, perform face detection on the historical image information to obtain face position information, and annotate the historical image information according to the face position information;

[0065] It should be noted that in order to construct a training data set, a large number of historical images are first obtained, and then face detection technology is used to detect each historical image, and then the location information of the facial part in each historical image is obtained, and then the facial part is annotated with a labeling box according to the location information of the facial part. The face detection methods generally include the FastR-CNN series, the cascade CNN series, and the SSD series.

[0066] Step S1022: cropping the historical image information according to the annotation results to obtain a target image containing the facial information of the person to be supervised, and extracting multiple facial key points in the target image, wherein the facial key points include the center of the eyes and the center of the lips;

[0067] The purpose of cropping the historical image is to focus the model learning on the face, thereby improving training accuracy. Furthermore, to reduce the impact of background information on the target image, the target image is also subjected to facial alignment. In this embodiment, three key points are first selected from the target image: the eye centers and the lip centers. The positional information of these three key points is then obtained.

[0068] Step S1023: creating a canvas and marking a plurality of target positions on the canvas, wherein the target positions correspond to the facial key points one-to-one, so as to map the pixel information of each facial key point to the corresponding target position one-to-one based on the position information of the binocular center and the position information of the lip center, thereby obtaining a new canvas on which the facial key points are mapped;

[0069] It should be noted that the monitoring management platform will call out a previously set blank canvas, and three target positions are marked on the blank canvas, which correspond to three facial key points respectively, and then the pixel information of these three facial key points is mapped to the corresponding target positions.

[0070] Step S1024: obtaining position information and pixel information of each facial feature in the target image, and calculating its mapping position information in the new canvas based on the position information of each facial feature in the target image;

[0071] It can be understood that after mapping the information of the above three facial key points onto the blank canvas, the position information and pixel information of the key points included in each facial feature are obtained, and the mapping position information of each point in each facial feature in the new canvas is calculated based on the affine transformation matrix.

[0072] Step S1025: mapping each facial feature in the target image to the new canvas one by one according to the mapping position information and pixel information of each facial feature, so as to obtain a face-aligned image;

[0073] It should be noted that the above-mentioned annotation and cropping can make the model learning object concentrated on the face part, and then perform face alignment operation on the target image containing the face part to reduce the interference of image background information and further improve the model recognition accuracy.

[0074] Step S1026: Obtain the pixel values ​​of all target points in the face alignment image, perform illumination compensation on each target point according to its pixel value, and input the face alignment image after illumination compensation into the initial face recognition model for training to obtain the final face recognition model.

[0075] For the stability of the algorithm model, in this embodiment, the image is also numerically normalized, and then the face alignment images obtained under different light intensities and different light source directions are compensated to reduce the image signal changes caused by lighting changes.

[0076] Specifically, the pixel values ​​of all target points are compensated according to the following formula:

[0077] X pt =(X pi -127.5) / 128

[0078] Among them, X pt Indicates the pixel value of the target point after illumination compensation, X pi Indicates the pixel value of the target point before illumination compensation.

[0079] It should also be noted that the network structure of deep convolutional neural networks is a key factor affecting the learning of deep facial features. The trend of face recognition methods based on deep learning is to use a single network. The three representative network architectures for deep face recognition are VGGNet, GoogLeNet and ResNet.

[0080] The core strategy of ResNet is to increase cross-layer connections and directly learn the residuals between layers. In this step, a residual module is used for deep learning. The input of the residual module is x, and the output is F(x)+x, where F(x) is the residual. The intermediate parameter layer only needs to learn the residual part, which can effectively reduce the training error. In addition, the cross-layer connection of this identity mapping avoids the vanishing gradient during the back-propagation process, which is conducive to training deeper networks.

[0081] Step S103: obtaining facial feature distance values ​​and facial contour feature distance values ​​of the facial image and the surveillance image;

[0082] It should be noted that after the model identifies and extracts the features of the surveillance image, it uses Euclidean distance or cosine similarity as a measure of feature distance. Assuming that the features of the face image xi and the surveillance image xj are f(xi) and f(xj) respectively, when the distance between the features is within the preset threshold τ, it is considered that the two images are from the same person.

[0083] Step S104: determining whether the facial feature distance value and the facial contour feature distance value are both less than a preset feature distance threshold;

[0084] Step S105: If it is determined that the facial feature distance value and the facial contour feature distance value are both less than a preset feature distance threshold, then the surveillance image identified in the current frame is determined to be consistent with the facial image of the person to be supervised, and the name of the target area corresponding to the surveillance image is obtained within a first preset time;

[0085] It can be understood that the first preset time is set to quickly trigger the retrieval of the corresponding target area name. The target area name is obtained based on the transmission object of the surveillance image, that is, each camera corresponds to the target area name. When the monitoring management platform knows that the currently analyzed surveillance image is consistent, it obtains the camera name corresponding to the surveillance image, and then obtains the corresponding target area name based on the camera name.

[0086] Step S106: numbering the surveillance images according to the target area name, and extracting the next frame of surveillance images from the surveillance video of the target space for repeated recognition;

[0087] It should be pointed out that in the process of numbering surveillance images, if the target area name is the same, it will be marked as the same batch, and the images of the same batch will be numbered according to the time information of the surveillance image, so that the surveillance images can be sorted in chronological order according to the size of the number, and the target video can be obtained.

[0088] Step S107: retrieve all surveillance images containing numbers, and sort and aggregate the retrieved surveillance images according to the numbering results to obtain the target video.

[0089] It can be understood that the target video produced includes multiple batches and the same batch is monitoring images of consecutive moments, which greatly facilitates the staff to view the batch videos of a certain target area at any time.

[0090] In summary, according to the above-mentioned monitoring data screening method, by automatically identifying all monitoring images to confirm whether the monitoring image is related to the person to be supervised, a monitoring video of the person to be supervised is accurately compiled, eliminating the need for manual screening and investigation of massive amounts of videos, which greatly reduces labor intensity. Specifically, the identity information of the person to be supervised is first obtained, and the facial image of the person to be supervised is retrieved from a preset image library based on the identity information. Then, all monitoring images of the target area are sequentially input into the trained recognition model to obtain multiple facial features contained in the monitoring image. The multiple facial features in the monitoring image are then compared with the multiple facial features in the facial images of the person to be supervised previously retrieved to determine whether the person in the monitoring image is the person to be supervised. If so, it indicates that the person to be supervised has passed through the target area. The names of all target areas that the person to be supervised has passed through in the target space are then obtained, and a monitoring video of the person to be supervised's behavior is generated, solving the problems of high manual labor intensity and easy information loss caused by manual data screening and investigation.

[0091] See also Figure 4 , which is a schematic diagram of the structure of a monitoring data screening system in a third embodiment of the present invention, is applied to a monitoring management platform and includes:

[0092] The identity information acquisition module 10 is used to obtain the identity information uploaded by the person to be supervised, and retrieve the facial image corresponding to the person to be supervised from the preset facial image database according to the identity information;

[0093] A facial feature recognition module 20 is configured to sequentially input surveillance images of a target area into a trained final facial recognition model to identify facial features contained in the surveillance images based on the final facial recognition model. The facial features include facial contour features and multiple facial features. The supervision space includes multiple target areas.

[0094] A face detection module 30 is used to determine whether the facial features of the face image corresponding to the person to be supervised are consistent with the facial features of the monitoring image;

[0095] Furthermore, the face detection module 30 further includes:

[0096] A feature distance acquisition unit, used to acquire the facial feature distance values ​​and the facial contour feature distance values ​​of the facial image and the surveillance image;

[0097] The feature distance detection unit is used to determine whether the facial feature distance value and the facial contour feature distance value are both less than a preset feature distance threshold.

[0098] The behavior monitoring video generation module 40 is used to determine that the person to be supervised has passed through the target area if the facial features contained in the facial image are consistent with the facial features contained in the monitoring image, and to summarize the names of all target areas passed by the person to be supervised, so as to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all target areas passed by the person to be supervised.

[0099] Furthermore, the behavior monitoring video generation module 40 also includes:

[0100] a target area name acquisition unit, configured to determine that the surveillance image identified in the current frame is consistent with the facial image of the person to be supervised if both the facial feature distance value and the facial contour feature distance value are determined to be less than a preset feature distance threshold, and to acquire the target area name corresponding to the surveillance image within a first preset time;

[0101] A surveillance image retrieval unit, configured to number the surveillance images according to the target area name, and extract the next frame of surveillance image from the surveillance video of the target space for repeated recognition;

[0102] The target video synthesis unit is used to retrieve all surveillance images containing numbers, and sort and summarize the retrieved surveillance images according to the numbering results to obtain the target video.

[0103] Furthermore, in some optional embodiments of the present invention, the system further includes:

[0104] a labeling module, configured to obtain a plurality of historical image information, perform face detection on the historical image information to obtain face position information, and label the historical image information according to the face position information;

[0105] A key point extraction module is used to crop the historical image information according to the annotation results to obtain a target image containing the facial information of the person to be supervised, and extract multiple facial key points in the target image, wherein the facial key points include the center of the binocular eyes and the center of the lips;

[0106] A key point mapping module is used to create a canvas and mark multiple target locations on the canvas. The target locations correspond to the facial key points one by one, so as to map the pixel information of each facial key point to the corresponding target location based on the position information of the binocular center and the position information of the lip center, thereby obtaining a new canvas with the facial key points mapped;

[0107] A mapping position information acquisition module is used to obtain the position information and pixel information of each facial feature in the target image, so as to calculate the mapping position information of each facial feature in the new canvas based on the position information of each facial feature in the target image;

[0108] a mapping execution module, configured to map each facial feature in the target image to the new canvas one by one according to mapping position information and pixel information of each facial feature, so as to obtain a face-aligned image;

[0109] A model training module is used to obtain the pixel values ​​of all target points in the face alignment image, perform illumination compensation on each target point according to the pixel value of each target point, and input the face alignment image after illumination compensation into the initial face recognition model for training to obtain the final face recognition model;

[0110] The pixel values ​​of all target points are compensated according to the following formula:

[0111] X pt =(X pi -127.5) / 128

[0112] Among them, X pt Indicates the pixel value of the target point after illumination compensation, X pi Indicates the pixel value of the target point before illumination compensation.

[0113] In summary, according to the above-mentioned monitoring data screening system, by automatically identifying all monitoring images to confirm whether the monitoring image is related to the person to be supervised, a monitoring video of the person to be supervised is accurately compiled, eliminating the need for manual screening and investigation of massive amounts of videos, which greatly reduces labor intensity. Specifically, the identity information of the person to be supervised is first obtained, and the facial image of the person to be supervised is retrieved from a preset image library based on the identity information. Then, all monitoring images of the target area are sequentially input into the trained recognition model to obtain multiple facial features contained in the monitoring image. The multiple facial features in the monitoring image are then compared with the multiple facial features in the facial images of the person to be supervised previously retrieved to determine whether the person in the monitoring image is the person to be supervised. If so, it means that the person to be supervised has passed through this target area. Then, the names of all target areas that the person to be supervised has passed through in the target space are obtained, and a monitoring video of the behavior of the person to be supervised is generated, solving the problems of high manual labor intensity and easy information loss caused by manual data screening and investigation.

[0114] Another aspect of the present invention further provides a readable storage medium having one or more programs stored thereon, which implement the above-mentioned monitoring data screening method when executed by a processor.

[0115] On the other hand, the present invention further provides a computer device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the above-mentioned monitoring data screening method.

[0116] Those skilled in the art will appreciate that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device), or in conjunction with such instruction execution system, apparatus, or device. For purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by an instruction execution system, apparatus, or device, or in conjunction with such instruction execution system, apparatus, or device.

[0117] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0118] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement the hardware: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0119] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0120] The above-described embodiments merely illustrate several embodiments of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A monitoring data screening method, applied to a monitoring management platform, characterized in that: The method comprises: Obtaining identity information uploaded by the person to be supervised, and retrieving a facial image corresponding to the person to be supervised from a preset facial image database based on the identity information; Acquire multiple copies of historical image information, perform face detection on the historical image information to obtain face position information, and annotate the historical image information according to the face position information. Crop the historical image information according to the annotated result to obtain a target image containing the face information of the person to be supervised, and extract multiple face key points in the target image. The face key points include the center of the binocular eyes and the center of the lips. Create a preset blank canvas and mark multiple target positions on the canvas. The target positions correspond to the face key points one by one. Map the pixel information of each face key point to the corresponding target position one by one according to the position information of the binocular eyes and the position information of the lip center to obtain a new canvas for mapping the face key points. Acquire the position information and pixel information of each face feature in the target image to obtain the face key points according to the target image. The position information of each facial feature in the image is calculated using an affine transformation matrix to obtain its mapping position information in the new canvas, and the facial features are mapped one by one to the new canvas according to the mapping position information and pixel information of each facial feature in the target image to obtain a face alignment image, and the pixel values ​​of all target points in the face alignment image are obtained to perform illumination compensation on each target point according to its pixel value, and the face alignment image after illumination compensation is input into the initial face recognition model for training to obtain a final face recognition model, and the monitoring image of the target area is sequentially input into the trained final face recognition model to identify the facial features contained in the monitoring image according to the final face recognition model, wherein the facial features include facial contour features and multiple facial features, and the supervision space includes multiple target areas; Determining whether the facial features of the facial image corresponding to the person to be supervised are consistent with the facial features of the monitoring image; If the facial features contained in the facial image are consistent with the facial features contained in the surveillance image, it is determined that the person to be supervised has passed through the target area, and the names of all the target areas passed by the person to be supervised are summarized to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all the target areas passed by the person to be supervised.

2. The monitoring data screening method according to claim 1, characterized in that: The step of determining whether the facial features contained in the facial image corresponding to the person to be supervised are consistent with the facial features contained in the monitoring image includes: Obtaining facial feature distance values ​​and facial contour feature distance values ​​of the face image and the surveillance image; Determine whether the facial feature distance value and the facial contour feature distance value are both less than a preset feature distance threshold.

3. The monitoring data screening method according to claim 2, characterized in that: If the facial features included in the facial image are consistent with the facial features included in the surveillance image, determining that the person to be supervised has passed through the target area, and summarizing the names of all the target areas passed by the person to be supervised to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all the target areas passed by the person to be supervised includes the following steps: If it is determined that the facial feature distance value and the facial contour feature distance value are both less than the preset feature distance threshold, then it is determined that the surveillance image recognized in the current frame is consistent with the facial image of the person to be supervised, and the name of the target area corresponding to the surveillance image is obtained within a first preset time; The surveillance images are numbered according to the target area name, and the next frame of surveillance images is extracted from the surveillance video of the target space for repeated recognition; All surveillance images containing numbers are retrieved, and the retrieved surveillance images are sorted and summarized according to the numbering results to obtain the target video.

4. The monitoring data screening method according to claim 1, characterized in that: The method further comprises: The pixel values ​​of all target points are compensated according to the following formula: X pt =(X pi -127.5) / 128 Among them, X pt Indicates the pixel value of the target point after illumination compensation, X pi Indicates the pixel value of the target point before illumination compensation.

5. A monitoring data screening system, applied to a monitoring management platform, characterized in that: The system comprises: An identity information acquisition module is used to acquire the identity information uploaded by the person to be supervised, and to retrieve a facial image corresponding to the person to be supervised from a preset facial image database based on the identity information; A facial feature recognition module is configured to sequentially input surveillance images of a target area into a trained final facial recognition model to identify facial features contained in the surveillance images based on the final facial recognition model. The facial features include facial contour features and multiple facial features. The supervision space includes multiple target areas. A face detection module is used to determine whether the facial features of the face image corresponding to the supervised person are consistent with the facial features of the monitoring image; The behavior monitoring video generation module is used to determine that the person to be supervised has passed through the target area if the facial features contained in the facial image are consistent with the facial features contained in the monitoring image, and to summarize the names of all target areas passed by the person to be supervised, so as to generate a behavior monitoring video corresponding to the person to be supervised based on the names of all target areas passed by the person to be supervised.

6. A readable storage medium, characterized in that: include: The readable storage medium stores one or more programs, which, when executed by a processor, implement the monitoring data screening method according to any one of claims 1 to 4.

7. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein: The memory is used to store computer programs; When the processor is used to execute the computer program stored in the memory, it implements the monitoring data screening method described in any one of claims 1-4.

Images