Campus video monitoring analysis system applied to smart school

Abstract

The invention relates to the technical field of monitoring equipment analysis, in particular to a campus video monitoring analysis system applied to a smart school, which comprises an image acquisition unit for acquiring a video image in a dormitory corridor and recording the video image as a monitoring video, and marking a plurality of rectangular areas in the monitoring video, each rectangular area corresponding to a dormitory door; the distinguishing and marking unit is used for identifying and analyzing the same human body contour in different pictures in the monitoring video and marking the human body contour in the monitoring picture; the state supervision unit constructs a rectangular plane coordinate system, draws a plurality of contour patterns corresponding to the target contour, constructs a target contour movement vector based on head center point coordinates of the contour patterns, marks the target contour as a target outgoing contour in combination with the position of the rectangular area, and marks a target dormitory; and based on the unit face database corresponding to the target dormitory, performing face recognition for the target going-out contour, and marking the corresponding student as a going-out state.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring device analysis, and particularly to a campus video monitoring and analysis system applied to a smart school. Background Art

[0002] The campus video monitoring system is an extremely important part of the construction of a smart campus. It deploys high-definition cameras in various key areas of the campus (such as teaching buildings, libraries, dormitories, campus entrances and exits, playgrounds, etc.) to achieve all-round and real-time dynamic monitoring of the campus environment. The video monitoring system also supports the remote viewing function, and management personnel can view the real-time campus images at any time through a computer or a mobile device, which is convenient for remote supervision and management of campus order.

[0003] Although the existing campus monitoring systems can achieve wide-range coverage monitoring and record the video data of students' daily behaviors, the video materials recorded by the systems are usually only saved as materials and used as important evidence for campus security incident investigations. It is impossible to record students' behaviors, especially the monitoring devices inside the dormitories, which are difficult to monitor and record students' going-out behaviors, resulting in an increase in the difficulty for the school authorities to regulate students' work and rest based on the monitoring devices. Summary of the Invention

[0004] In view of the above-mentioned drawbacks of the existing technology, the present invention provides a campus video monitoring and analysis system applied to a smart school, which can effectively solve the problem that it is difficult for the monitoring devices inside the dormitories in the existing technology to monitor and record students' going-out behaviors.

[0005] To achieve the above object, the present invention is realized through the following technical solutions:

[0006] The present invention provides a campus video monitoring and analysis system applied to a smart school, which at least includes:

[0007] An image acquisition unit, where monitoring devices are arranged at multiple positions in the dormitory corridor to obtain the video images in the dormitory corridor, which are recorded as monitoring videos, and multiple rectangular areas are marked in the monitoring videos, and each rectangular area corresponds to a dormitory door;

[0008] A discrimination and marking unit, which presets an analysis period, obtains the target video within the most recent analysis period and performs frame extraction processing, screens out the analysis pictures and identifies the human contours therein, and analyzes multiple consecutive analysis pictures with the number of human contours greater than 0:

[0009] Extract any two adjacent analysis pictures and record them as comparison pictures and perform overlapping comparison. Based on the overlapping area of the human contours and the position of the center point of the head region in the two comparison pictures, the corresponding two human contours in the two comparison pictures are recorded as a group of the same contour;

[0010] Label the human body contours in the analysis screen, and keep the labels of the human body contours with the same contour in different analysis screens consistent;

[0011] The state supervision unit divides the human body contours into different target contours based on their serial numbers and performs analysis on each target contour:

[0012] Construct a plane rectangular coordinate system and draw multiple contour patterns corresponding to the target contour, construct the target contour movement vector based on the head center point coordinates of the contour pattern, mark the target contour as the target outgoing contour and mark the target dormitory in combination with the rectangular area position;

[0013] Based on the unit face database corresponding to the target dormitory, face recognition is performed on the target out-of-town outline, and the corresponding students are marked as out-of-town.

[0014] Furthermore, the frame extraction process is as follows:

[0015] An analysis cycle is preset, a target video in the most recent analysis cycle is obtained and split into multiple unit frame images, and the total number of unit frame images is equal to the refresh rate of the target video multiplied by the analysis cycle;

[0016] A plurality of equally spaced unit frame images are extracted from a plurality of unit frame images and recorded as analysis images. The number of unit frame images spaced between the analysis images is equal to the interval duration multiplied by the refresh rate of the target video, wherein the interval duration is a preset value.

[0017] Furthermore, the overlap comparison process is as follows:

[0018] Draw all the human body contours in the comparison picture, obtain the center point of the head area of ​​each human body contour as the head positioning point, and each head positioning point corresponds to a human body contour;

[0019] The head positioning points in each comparison picture are sorted and numbered as Where j = A or B, representing the previous comparison picture and the next comparison picture respectively, i is the serial number of each head positioning point, and the head positioning point The corresponding human body contour is denoted as

[0020] Overlap the two comparison images and extract any two head positioning points Compare, x, y are constant values ​​not greater than the number of human body contours in the corresponding comparison picture, calculate the distance between the two and record it as the first judgment value P, extract the head positioning point Corresponding human body contour For overlap comparison, use the formula The second judgment value S is calculated, where S1 and S2 respectively represent the area of the human body contour and S3 represents the area of the overlapping part of the human body contour .

[0021] Furthermore, the process of analyzing the labeling of the human body contours in the picture is as follows:

[0022] The first judgment value and the second judgment value are obtained and substituted into the formula PD = η1*P + η2*S for calculation to obtain the same judgment value PD, where η1 and η2 are both preset weight coefficients;

[0023] There is a preset same judgment threshold. When the same judgment value is greater than the same judgment threshold, the corresponding two human body contours are recorded as a group of the same contour. On the premise of maintaining the consistency of the serial numbers of the same contour, the human body contours in the subsequent comparison pictures are re-numbered.

[0024] Furthermore, the process of constructing the movement vector is as follows:

[0025] Multiple contour patterns corresponding to the same target contour are obtained. Each contour pattern corresponds to an analysis picture, and each contour pattern is sorted and numbered as PICTURE n according to the shooting time sequence of the analysis pictures, where n is the serial number of the contour pattern, n = 1, 2, 3,..., m, and m is the total number of contour patterns;

[0026] A plane rectangular coordinate system is constructed, and the head center points corresponding to each contour pattern are plotted in the plane rectangular coordinate system and recorded as head coordinate points. The coordinates of each head coordinate point are recorded as (X n , Y n ). The coordinates of the two head coordinate points (X1, Y1) and (X m , Y m ) corresponding to the contour pattern PICTURE1 and the contour pattern PICTUREm are respectively extracted. Based on the coordinates of the two head coordinate points (X1, Y1) and (X m , Y m ), a movement vector is constructed

[0027] Furthermore, the process of determining the target outer contour and the target dormitory is as follows:

[0028] The distances between the coordinates of the two head coordinate points (X1, Y1) and (X m , Y m ) and the nearest rectangular area are respectively calculated and recorded as the analysis distances. The rectangular area corresponding to the smaller value of the two analysis distances is selected as the target rectangle, and a straight line parallel to the corridor direction is drawn through the center point of the target rectangle and recorded as the target line;

[0029] When the movement vector points to the target line, the target contour is denoted as the target outgoing contour, and the dormitory corresponding to the target rectangle is denoted as the target dormitory.

[0030] Furthermore, the construction process of the unit face database corresponding to the target dormitory is as follows:

[0031] A unit face database is constructed for each dormitory door. The unit face database contains multiple student face data belonging to the corresponding dormitory. When a student checks into the dormitory, the face data is entered and added to the unit face database. When a student moves out of the dormitory, the student's face data is removed from the unit face database. The unit face database is bound to the rectangular area, and the unit face database corresponding to the target dormitory is obtained and denoted as the target unit library.

[0032] Furthermore, the process of marking the student's outgoing status is as follows:

[0033] The area value of each contour pattern is obtained and denoted as the contour area. Each contour pattern is grayscale processed to obtain a contour grayscale image. There is a preset skin color grayscale interval. The proportion of pixel points within the skin color grayscale interval in the contour grayscale image is obtained and denoted as the skin color proportion. The skin color proportion FS and the contour area MJ are extracted and substituted into the formula LK = λ1 * FS + λ2 * MJ for calculation to obtain the contour judgment value LK, where λ1 and λ2 are both preset weight coefficients;

[0034] The head region in the contour pattern with the largest contour judgment value is extracted and denoted as the comparison region. The image within the comparison region is intercepted and compared one by one with the face data in the target unit library to calculate the similarity. The face data with the largest similarity is bound to the target contour corresponding to the contour pattern, and the student corresponding to the face data is marked as being in the outgoing status.

[0035] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above-mentioned system is implemented.

[0036] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned system is implemented.

[0037] The technical solution provided by the present invention has the following beneficial effects compared with the known prior art:

[0038] 1. The present invention overlaps and compares two adjacent analysis screens, analyzes and calculates the displacement distance and overlapping area of the head center points between multiple sets of human contours in the two analysis screens, and then determines whether two human contours are the same contour. Further, through continuous comparative analysis, a set of continuous identical contours can be determined, so as to continuously mark the human contours in the analysis screen, and then distinguish different human contours, so that each serial number preferably refers to the same human contour in different analysis screens, which is convenient for subsequent differential analysis of human contours, and overcomes the problem that although the human contour recognition algorithm in the prior art can recognize human contours, it cannot distinguish and continuously record different human contours.

[0039] 2. The present invention sorts and numbers multiple contour patterns corresponding to the same contour, so as to perform time-series analysis on the contour patterns. A moving vector is constructed to analyze the moving direction of the person corresponding to the target contour, and the current moving trajectory direction of the person is judged according to the direction of the moving vector. Further, based on the moving trajectory direction of the person and combined with the straight-line position of the corresponding dormitory door, it can be judged whether the person enters or leaves the dormitory. Combining the unit information library corresponding to each dormitory, face data comparison is performed on the human contour corresponding to the person going out, and the student corresponding to the target contour can be confirmed, ensuring that the outgoing image of the student can be captured, so as to monitor the safety of the student. And the data capacity of the unit information library is small, which can save the time required for one-by-one comparison in the comparison process and improve the recognition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is the overall module block diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0043] The following further describes the present invention with reference to the embodiments.

[0044] See also Figure 1 , the campus video surveillance analysis system applied to smart schools at least includes:

[0045] The image acquisition unit is provided with monitoring equipment at multiple locations in the dormitory corridor, and is used to acquire video images of the dormitory corridor. The video image data of students entering and leaving the dormitory is recorded by the monitoring equipment. The video images acquired by monitoring in the dormitory corridor are recorded as monitoring videos. Multiple rectangular areas are marked in the monitoring videos, each of which corresponds to a dormitory door. The image in the rectangular area covers the area where the dormitory door is located. That is to say, the image change in the rectangular area usually corresponds to the state change of the dormitory door.

[0046] The distinguishing and marking unit identifies the human body contours in the corridor image and distinguishes and marks them, giving a distinguishing serial number to different human body contours. The distinguishing and marking process is as follows:

[0047] Step 1: An analysis cycle is preset (in a specific embodiment, the analysis cycle is 3 seconds long), the target video in the most recent analysis cycle is obtained and split into multiple unit frame images, the total number of unit frame images is equal to the refresh rate of the target video multiplied by the analysis cycle, and multiple unit frame images with equal intervals are extracted from the target video as analysis images. The number of unit frame images between analysis images is a preset value, which is set by the staff according to the needs of specific implementation;

[0048] It should be noted that the target video is composed of multiple continuously shot unit frames. By sampling frames at equal intervals and filtering out unit frames with highly similar content, the number of analyzed frames can be reduced to improve the efficiency of image analysis.

[0049] Step 2: respectively obtain the number of human body contours in each analysis picture, record the analysis picture with the number of human body contours greater than 0 as a dynamic picture, and record the analysis picture with the number of human body contours equal to 0 as a static picture. When the analysis picture changes from a static picture to a dynamic picture, continuously analyze the dynamic picture;

[0050] It should be noted that the recognition of human body contours mainly relies on the image recognition algorithm based on deep learning in the prior art, which can use the deep neural network to learn the human body contour features in the picture, so as to automatically recognize the shape of the human body contour in the video picture and depict and mark it with a corresponding closed pattern after recognizing the human body contour. I will not go into details here. By distinguishing between static pictures and dynamic pictures, it is helpful to determine the initial picture of human body movement in the monitoring picture, so as to conduct continuous analysis based on the initial picture, and then distinguish different human body contours. Although the human body contour recognition algorithm in the prior art can recognize the human body contour, it cannot distinguish different human body contours.

[0051] Step 3: Extract any two adjacent analysis screens for overlapping comparison. The comparison process is as follows:

[0052] Denote the two adjacent analysis screens as comparison screens, draw all the human outlines in the comparison screens, and obtain the center points of the head regions of each human outline, which are denoted as head positioning points. The head positioning points are used to determine the head positions of each human outline, and each head positioning point corresponds to a human outline;

[0053] Sort and label the head positioning points in each comparison screen respectively as where j = A or B, representing the prior comparison screen and the subsequent comparison screen respectively, and i is the serial number of each head positioning point (sorted in the order of 1, 2, 3...). Denote the human outline corresponding to the head positioning point as

[0054] Overlap the two comparison screens, extract any two head positioning points for comparison. x and y are constant values not greater than the number of human outlines in the corresponding comparison screens. Calculate the distance value between them, which is denoted as the first judgment value P. Extract the human outline corresponding to the head positioning point as for overlapping comparison, and calculate the second judgment value S through the formula where S1 and S2 respectively represent the areas of the human outlines and S3 represents the area of the overlapping part of the human outline ;

[0055] When a group of human outlines in the two comparison screens highly overlap and the distance between the head center points is relatively close, it can generally be considered that this group of human outlines are the different outline positions of the same person in the two screens. Because when a person is in a moving state, in two pictures taken by the same monitoring device at different times, the human outline of this person will have a displacement change in the screen. However, due to the shooting interval limitation, the moving distance of the human head and the offset of the human outline will be within a certain range. Therefore, it can be used as the judgment basis for whether a group of human outlines in the two comparison screens are the corresponding outlines of the same person.

[0056] Step 4: Obtain the first judgment value and the second judgment value and substitute them into the formula PD = η1*P + η2*S for calculation to obtain the same judgment value PD, where η1 and η2 are both preset weight coefficients, which are specifically set by the staff during the specific implementation process. There is a preset same judgment threshold. When the same judgment value is greater than the same judgment threshold, the corresponding two human silhouettes are recorded as a group of the same silhouette. On the premise of maintaining the consistency of the serial numbers of the same silhouette (that is, making the serial numbers of the two same silhouettes equal), re-number the human silhouettes in the subsequent comparison pictures.

[0057] It should be noted that by continuously comparing two adjacent analysis pictures, the human silhouettes in the analysis pictures are continuously marked, and then different human silhouettes are distinguished, so that each serial number can preferably refer to the same human silhouette in different analysis pictures, which is convenient for subsequent differential analysis of the human silhouettes.

[0058] The status supervision unit distinguishes them based on the serial numbers of the human silhouettes, screens out the students going out in combination with the movement trajectories of the human silhouettes, and performs face recognition on the students going out to determine the data information of the students going out, where:

[0059] The human silhouettes corresponding to different serial numbers are recorded as target silhouettes.

[0060] It should be noted that the target silhouette refers to the human silhouette with a serial number. Each target silhouette corresponds to the same silhouette in multiple analysis pictures. That is to say, the difference between the human silhouette and the target silhouette is that the target silhouette is a specific silhouette, while the human silhouette generally refers to any human silhouette recognized by the recognition algorithm. Therefore, the target silhouette does not refer to a certain silhouette pattern, but multiple silhouette patterns. These silhouette patterns generally have differences in size and shape due to different recording times.

[0061] Analyze each target silhouette:

[0062] S1: Obtain multiple silhouette patterns corresponding to the same target silhouette (that is, the silhouette patterns of the same person in the monitoring pictures at different times). Each silhouette pattern corresponds to an analysis picture. Sort and number each silhouette pattern according to the shooting time sequence of the analysis pictures and record it as PICTURE n , n is the serial number of the silhouette pattern, n = 1, 2, 3,..., m, m is the total number of silhouette patterns. When n = 1, it means that the target silhouette first appears in the analysis picture;

[0063] By sorting and numbering the silhouette patterns, time series analysis can be performed on the silhouette patterns, and thus based on the change characteristics of the silhouette patterns, the movement trend of the person corresponding to the target silhouette can be determined.

[0064] S2: Construct a plane rectangular coordinate system, plot the head center point corresponding to each contour pattern in the plane rectangular coordinate system and denote it as the head coordinate point, and obtain the coordinates of each head coordinate point as (X n , Y n ). Respectively extract the coordinates of the two head coordinate points (X1, Y1) and (X m ) corresponding to the contour pattern PICTURE1 and the contour pattern PICTURE m , Y m ). Based on the coordinates of the two head coordinate points (X1, Y1) and (X m , Y m ), construct a movement vector

[0065]

[0066] S3: Calculate the distances between the coordinates of the two head coordinate points (X1, Y1) and (X m , Y m ) and the nearest rectangular area respectively, and denote them as the analysis distances. Select the rectangular area corresponding to the smaller value among the two analysis distances as the target rectangle. Draw a straight line parallel to the corridor direction through the center point of the target rectangle and denote it as the target line. When the movement vector points to the target line, denote the target contour as the target outgoing contour, and denote the dormitory corresponding to the target rectangle as the target dormitory;

[0067] By constructing a movement vector to analyze the movement direction of the person corresponding to the target contour, and judging the current movement trajectory direction of the person according to the direction of the movement vector. Further, based on the movement trajectory direction of the person and combined with the straight line where the corresponding dormitory door is located, it is possible to judge whether the person is entering or leaving the dormitory.

[0068] S4: Construct a unit face database for each dormitory door. The unit face database contains multiple student face data belonging to the corresponding dormitory. When a student checks into the dormitory, the face data is entered and added to the unit face database. When a student moves out of the dormitory, the student's face data is removed from the unit face database. Bind the unit face database to the rectangular area, and obtain the unit face database corresponding to the target dormitory and denote it as the target unit library;

[0069] By constructing multiple unit information databases with small capacities, the monitoring device can compare the data capacity of the database based on the video recognition area limit, thereby saving the time required for one-by-one comparison during the comparison process and improving the recognition efficiency. Compared with the total database comparison method in the prior art, it is more convenient and fast.

[0070] S5: Obtain the area value of each contour pattern, denoted as the contour area. Perform grayscale processing on each contour pattern to obtain a contour grayscale image. Preset a skin color grayscale interval (i.e., the range interval of the grayscale values of skin color pixel points after grayscale processing). Obtain the proportion of pixel points within the skin color grayscale interval in the contour grayscale image, denoted as the skin color proportion. Extract the skin color proportion FS and the contour area MJ, substitute them into the formula LK = λ1 * FS + λ2 * MJ for calculation to obtain the contour judgment value LK, where λ1 and λ2 are both preset weight coefficients;

[0071] The contour judgment value comprehensively reflects the distinguishability of the target contour in the analysis picture collected by the monitoring device. This value depends on the overall size of the human body contour (usually, the larger the human body contour, the clearer the head pattern, and the more convenient for comparison and analysis) and the proportion of the skin color area (the proportion of the skin color area reflects sideways whether the human body contour is facing or side-facing the monitoring device. The larger the proportion of the skin color area, the clearer the facial area is captured, and the more convenient for comparison and analysis).

[0072] S6: Extract the head area in the contour pattern with the largest contour judgment value as the comparison area. Intercept the image within the comparison area and compare it with the face data in the target unit library one by one to calculate the similarity. Select the face data with the largest similarity and bind it to the target contour corresponding to this contour pattern. Mark the student corresponding to this face data as the going-out status.

[0073] It should be noted that through face data comparison, the student corresponding to the target contour is further confirmed, so as to achieve the final information confirmation for each human body contour that meets the specific going-out image data characteristics. Furthermore, it is ensured that the going-out images of students can be captured through this video monitoring, so as to determine the going-out time of students, and based on the going-out time, monitor the safety of students, or manage and supervise the living habits of students. For example, prevent students from staying in the dormitory for a long time without exercise.

[0074] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above system is implemented.

[0075] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the above system is implemented.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. The campus video surveillance analysis system applied to smart schools is characterized by: include: The image acquisition unit is provided with monitoring devices at multiple locations in the dormitory corridor, and the video images in the dormitory corridor are obtained and recorded as monitoring videos, and multiple rectangular areas are marked in the monitoring videos, each of which corresponds to a dormitory door; The distinguishing marking unit has a preset analysis cycle. The target video in the most recent analysis cycle is obtained and frame extraction is performed. The analysis screen is screened out and the human contours therein are identified. Multiple continuous analysis screens with a number of human contours greater than 0 are analyzed: Extract any two adjacent analysis images as comparison images and perform overlapping comparison. Based on the overlapping area of ​​the human body contours in the two comparison images and the position of the center point of the head area, record the two corresponding human body contours in the two comparison images as a group of the same contours; Label the human body contours in the analysis screen, and keep the labels of the human body contours with the same contour in different analysis screens consistent; The state supervision unit divides the human body contours into different target contours based on their serial numbers and performs analysis on each target contour: Construct a plane rectangular coordinate system and draw multiple contour patterns corresponding to the target contour, construct the target contour movement vector based on the head center point coordinates of the contour pattern, mark the target contour as the target outgoing contour and mark the target dormitory in combination with the rectangular area position; Based on the unit face database corresponding to the target dormitory, face recognition is performed on the target out-of-town outline, and the corresponding students are marked as out-of-town.

2. The campus video surveillance and analysis system applied to a smart school according to claim 1 is characterized in that: The frame extraction process is as follows: An analysis cycle is preset, a target video in the most recent analysis cycle is obtained and split into multiple unit frame images, and the total number of unit frame images is equal to the refresh rate of the target video multiplied by the analysis cycle; A plurality of equally spaced unit frame images are extracted from a plurality of unit frame images and recorded as analysis images. The number of unit frame images spaced between the analysis images is equal to the interval duration multiplied by the refresh rate of the target video, wherein the interval duration is a preset value.

3. The campus video surveillance and analysis system applied to a smart school according to claim 2 is characterized in that: The overlap comparison process is as follows: Draw all the human body contours in the comparison picture, obtain the center point of the head area of ​​each human body contour as the head positioning point, and each head positioning point corresponds to a human body contour; The head positioning points in each comparison picture are sorted and numbered as Where j = A or B, representing the previous comparison picture and the next comparison picture respectively, i is the serial number of each head positioning point, and the head positioning point The corresponding human body contour is denoted as Overlap the two comparison images and extract any two head positioning points Compare, x, y are constant values ​​not greater than the number of human body contours in the corresponding comparison picture, calculate the distance between the two and record it as the first judgment value P, extract the head positioning point Corresponding human body contour For overlap comparison, use the formula The second judgment value S is calculated, where S1 and S2 represent the human body contours respectively. The area of ​​​​the human body, S3 represents the human body contour The area of ​​the overlapping part.

4. The campus video surveillance and analysis system applied to a smart school according to claim 1 is characterized in that: The process of labeling the human body contour in the analysis picture is as follows: Obtain the first judgment value and the second judgment value and substitute them into the formula PD=η1*P+η2*S for calculation to obtain the same judgment value PD, where η1 and η2 are both preset weight coefficients; The same judgment threshold is preset. When the same judgment value is greater than the same judgment threshold, the corresponding two human body contours are recorded as a group of the same contours. On the premise of maintaining the consistency of the sequence number of the same contour, the human body contours in the subsequent comparison screen are reordered and labeled.

5. The campus video surveillance and analysis system applied to a smart school according to claim 1 is characterized in that: The motion vector construction process is as follows: Get multiple contour patterns corresponding to the same target contour, each contour pattern corresponds to an analysis screen, and sort each contour pattern according to the shooting time of the analysis screen and label it as PICTURE n , n is the serial number of the contour pattern, n = 1, 2, 3, ..., m, m is the total number of contour patterns; Construct a plane rectangular coordinate system, draw the head center point corresponding to each contour pattern in the plane rectangular coordinate system and record it as the head coordinate point, obtain the coordinates of each head coordinate point and record it as (X n ,Y n ), respectively extract the coordinates of the two head points (X1, Y1) and (X m ,Y m ), based on the coordinates of the two head points (X1, Y1), (X m ,Y m ) Construct the motion vector 6. The campus video surveillance and analysis system applied to a smart school according to claim 5 is characterized in that: The process of determining the target outgoing outline and target dormitory is as follows: Calculate the coordinates of the two head points (X1, Y1), (X m ,Y m ) The distance between the two nearest rectangular areas is recorded as the analysis distance, the rectangular area corresponding to the smaller value of the two analysis distances is selected as the target rectangle, and a straight line parallel to the corridor direction through the center point of the target rectangle is recorded as the target straight line; When the moving vector points to the target straight line, the target contour is recorded as the target outgoing contour, and the dormitory corresponding to the target rectangle is recorded as the target dormitory.

7. The campus video surveillance and analysis system applied to a smart school according to claim 6 is characterized in that: The process of constructing the unit face database corresponding to the target dormitory is as follows: A unit face database is constructed for each dormitory door. The unit face database contains multiple student face data belonging to the corresponding dormitory. When a student moves into the dormitory, the face data is entered and added to the unit face database. When the student moves out of the dormitory, the student's face data is removed from the unit face database. The unit face database is bound to the rectangular area, and the unit face database corresponding to the target dormitory is obtained and recorded as the target unit library.

8. The campus video surveillance and analysis system applied to a smart school according to claim 7 is characterized in that: The process of marking a student's out-of-office status is as follows: The area value of each contour pattern is obtained and recorded as the contour area, and each contour pattern is gray-scaled to obtain a contour grayscale image. A skin color grayscale interval is preset, and the proportion of pixels in the contour grayscale image located in the skin color grayscale interval is obtained and recorded as the skin color proportion. The skin color proportion FS and the contour area MJ are extracted and substituted into the formula LK=λ1*FS+λ2*MJ for calculation to obtain the contour judgment value LK, where λ1 and λ2 are both preset weight coefficients; The head area in the contour pattern with the largest contour judgment value is extracted and recorded as the comparison area. The image in the comparison area is intercepted and compared with the face data in the target unit library one by one to calculate the similarity. The face data with the largest similarity is selected and bound to the target contour corresponding to the contour pattern, and the student corresponding to the face data is marked as out.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the system according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the system according to any one of claims 1 to 8 is implemented.

Images