A method for identity analysis based on human gait video

Abstract

The present application relates to the technical field of target tracing, and particularly relates to an identity analysis method based on human gait video, comprising the following steps: step one, establishing a human gait feature database, calculating the acceleration of the key joint of personnel after labeling according to the image data of four groups of cameras, and establishing the human gait feature database; step two, acquiring any monitoring video in a target area, and calculating the acceleration of the key joint in each gait of the target personnel in the coordinate system of the camera, comparing with the gait feature database, and specially marking the first few with higher similarity; step three, manually excluding and confirming the identity. The acceleration of the key joint during continuous walking is solved, which can be identified as the unique gait feature of each personnel, thereby assisting the staff to confirm the identity of the personnel in the video data, determine the target personnel, and determine the personnel in close contact with the target personnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of target tracing, and particularly relates to an identity analysis method based on human gait video. BACKGROUND

[0002] In order to ensure the normal and safe operation of social work, the personnel flow in the target area is very important to monitor. The existing monitoring means is mostly to cover the target area without dead angle through multiple monitoring cameras to realize the personnel flow monitoring in the target area. However, this means relies on the recognition of the facial features of the personnel to confirm the identity, which has great limitations, especially for the identity recognition of the personnel wearing masks, which brings hidden dangers to the target tracing and the normal and safe operation of social work. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides an identity analysis method based on human gait video, which analyzes the human gait features of the video data in the prior art to assist the staff in confirming the identity of the personnel in the video data, determining the target personnel, and confirming the personnel in close contact with the target personnel, especially assisting in confirming the identity of the target personnel wearing masks, thereby quickly tracing and determining the trajectory, and ensuring the normal and safe operation of social work.

[0004] To achieve the above purpose, the present application adopts the following technical scheme:

[0005] The present application provides an identity analysis method based on human gait video, comprising the following steps:

[0006] Step one, establish a human gait feature database, the specific steps are as follows:

[0007] Step 1.1. Set up a straight route, and set up four groups of cameras to establish a space reference coordinate system, wherein two groups of the cameras are symmetrically arranged on the two sides of the route, and the camera optical axes of the cameras are perpendicular to the forward direction of the route, and the other two groups of the cameras are arranged at the starting point and the ending point of the route, and the camera optical axes are parallel to the route, and the space reference coordinate system takes the starting point of the route as the origin, and takes the direction along the route, the direction perpendicular to the route, and the direction perpendicular to the route as the space right-angle coordinate axes.

[0008] Step 1.2. instruct the personnel to walk straight from the starting point to the ending point on the route, and then mark the joint points of each step of the personnel;

[0009] Step 1.3. Calculate the coordinates of the labeled joints in the spatial reference coordinate system according to the image data of the four groups of cameras, and then calculate the horizontal and vertical accelerations of the joints in each step of the gait of the person walking continuously in the spatial reference coordinate system.

[0010] Step 1.4. Take the horizontal and vertical accelerations of the joints in the spatial reference coordinate system as gait features, and bind the gait features with the person's identity, then enter the database to establish the human gait feature database.

[0011] Step two, process the video to be identified, the specific steps are as follows:

[0012] Step 2.1. Obtain any monitoring video in the target area, and obtain the parameters of the camera corresponding to the monitoring video, and establish the coordinate system of the camera.

[0013] Step 2.2. Label the joints of each step of the target person in the monitoring video obtained in step 2.1, and calculate the coordinates of the joints in the coordinate system in each step of the gait of the target person, and then calculate the horizontal and vertical accelerations of the joints in the coordinate system when walking continuously.

[0014] Step 2.3. Compare the horizontal and vertical accelerations of the joints in the coordinate system obtained in step 2.2 with the human gait feature database, and rank according to the similarity, and specially mark the target person with higher similarity.

[0015] Step three, identify the identity of the target person specially marked in step two by manual.

[0016] Specifically, in step one, the height, gender, body shape, and personal habitual action of the person are also bound with the person's identity.

[0017] Specifically, in step three, the specific steps of manual identity recognition are as follows:

[0018] Step 3.1. Identify the gender of the specially marked target person, and exclude those who do not meet the gender requirements.

[0019] Step 3.2. Identify the height of the specially marked target person, and exclude those who do not meet the height requirements.

[0020] Step 3.3. Identify the body shape of the specially marked target person, and exclude those who do not meet the body shape requirements.

[0021] Step 3.4. Identify the personal habitual action of the target person marked specially, and confirm the identity of the target person in combination with the personal habitual action.

[0022] Specifically, in the step two, the obtained monitoring video should meet the condition of containing at least 5 continuous normal walking steps of the target person.

[0023] Specifically, in the step two, the joint points of the target person in the monitoring video are marked with a frame rate period of 15-30 FPS.

[0024] Specifically, the joint points include wrist joint, elbow joint, shoulder joint, neck joint, ankle joint, knee joint and hip joint.

[0025] Specifically, in the step two, a coordinate system is established, the road plane is a triangular plane ABU, ABCD is a trapezoidal area on the road plane shot by the camera, CD intersects with BU at C point, CD intersects with AU at D point, CD is parallel to AB, O point is the lens center point of the camera, OG is the optical axis of the camera, G point is the intersection of the optical axis of the camera and the road plane, and is also the intersection point of the diagonal of the field of view trapezoid ABCD, I point is the vertical projection of O point on the road plane, the height OI of the camera is h, UE intersects with AB at E point and is perpendicular to AB, UE intersects with CD at F point, JK intersects with UE at G point, JK intersects with AU at K point, JK intersects with BU at J point, JK is parallel to AB, in the road plane coordinate system, G point is defined as the origin of the coordinate system, UE is defined as the Y axis of the coordinate system, the forward direction of the person is defined as the direction of Y axis, and JK is defined as the X axis of the coordinate system.

[0026] G, A, B, C and D points of the road plane correspond to g, a, b, c and d points of the image plane respectively, a, b, c and d are four end points of the rectangle, g is the midpoint of the image plane, ef intersects with ab at e point, ef intersects with cd at f point, jk intersects with ef at g point, jk intersects with bc at j point, jk intersects with ad at k point, E, F, J and K points of the road plane correspond to e, f, j and k points of the image plane respectively, the height cb of the image plane is H, the width cd of the image plane is W, g point of the image plane is defined as the origin of the image plane coordinate system, ef is defined as the y axis of the image plane coordinate system, the forward direction of the person is defined as the direction of y axis, and jk is defined as the x axis of the image plane coordinate system, a point P on the road plane is taken, the coordinates of P point in the road plane coordinate system are (X p , Y p ), P point corresponds to p point in the image plane, and the coordinates of p point in the image plane coordinate system are (x p , y p ).

[0027] The plane perpendicular to the road plane and the camera optical axis is OEI, ML intersects with OG at G point and is perpendicular, ML intersects with the extension line of fF at L point, p y P represents the projection of p point on the image plane on ef, P y Z point is the projection of P point on the Y axis, and Z point is the projection of P point on the Y axis y P y The intersection point of ML, the horizontal field of view angle 2∠BUE of the camera is 2β0, the vertical field of view angle 2∠EOG is 2α0, the pitch angle ∠GOI is γ0, and ∠P y OG is α, and ∠P y OP is α1;

[0028] The imaging model in the Y axis direction is obtained according to the trigonometric function relationship and similar triangles, ;

[0029] Further ;

[0030] Further ;

[0031] The imaging model in the X axis direction is P x The projection point of P point on the X axis is in triangle △OP y L, ;

[0032] According to the similar triangle of P point imaging, ;

[0033] Finally .

[0034] The application also provides an identity analysis method based on human gait video, comprising the following steps:

[0035] Step one, since the movements of the joints of a person are not the same when walking, the movements of the joints are identified as gait characteristics unique to each person, thereby establishing a human gait characteristic database;

[0036] Step two, any monitoring video in a target area is obtained, and the horizontal and vertical accelerations of each joint in each gait of the target person in the camera coordinate system corresponding to the monitoring video are calculated and used for comparison with the human gait characteristic database, and the top several target persons with higher similarity are specially marked;

[0037] Step three, artificial exclusion confirms identity, by solving the accelerations of each joint when continuously walking, the accelerations can be identified as gait characteristics unique to each person, thereby assisting the staff to confirm the identity of the target person in the monitoring video, determining the confirmed person, and thereby determining the close contact person.

[0038] Specifically, in the step two, a coordinate system is established, the road plane is a triangular plane ABU, ABCD is a trapezoidal area on the road plane shot by the camera, CD intersects with BU at C point, CD intersects with AU at D point, CD is parallel to AB, O point is the lens center point of the camera, OG is the optical axis of the camera, G point is the intersection of the optical axis of the camera and the road plane, and is also the intersection of the diagonal of the field of view trapezoid ABCD, I point is the vertical projection of O point on the road plane, the height OI of the camera is h, UE intersects with AB at E point, and UE is perpendicular to AB, UE intersects with CD at F point, JK intersects with UE at G point, JK intersects with AU at K point, JK intersects with BU at J point, JK is parallel to AB, in the road plane coordinate system, G point is defined as the coordinate system origin, UE is defined as the coordinate system Y axis, the personnel advancing direction is defined as the Y axis direction, and JK is defined as the coordinate system X axis;

[0039] G, A, B, C, D points of the road plane correspond to g, a, b, c, d points of the image plane respectively, a, b, c, d are four end points of the rectangle, g is the midpoint of the image plane, ef intersects with ab at e point, ef intersects with cd at f point, jk intersects with ef at g point, jk intersects with bc at j point, jk intersects with ad at k point, E, F, J, K points of the road plane correspond to e, f, j, k points of the image plane respectively, the height cb of the image plane is H, the width cd of the image plane is W, g point of the image plane is defined as the coordinate origin of the image plane coordinate system, ef is defined as the y axis of the image plane coordinate system, the personnel advancing direction is defined as the y axis direction, and jk is defined as the x axis of the image plane coordinate system, a point P on the road plane is taken, the coordinates of P point in the road plane coordinate system are (X p , Y p ), P point corresponds to p point on the image plane, the coordinates of p point in the image plane coordinate system are (x p , y p );

[0040] The plane perpendicular to the optical axis of the camera and the road plane is OEI, ML intersects with OG at G point and is perpendicular, ML intersects with the extension line of fF at L point, p y represents the projection of p point on the image plane on ef, P y point is the projection of P point on the road plane on Y axis, Z point is the intersection of p y P y and ML, the horizontal field of view angle of the camera 2∠BUE is 2β0, the vertical field of view angle 2∠EOG is 2α0, the pitch angle ∠GOI is γ0, ∠P y OG is α, and ∠P y OP is α1;

[0041] The imaging model in Y-axis direction can be obtained according to trigonometric function relation and similar triangle, ;

[0042] Further ;

[0043] Further ;

[0044] The imaging model in X-axis direction, P x is the projection point of P point on X-axis, in triangle △OP y L, ;

[0045] According to the similar triangle of P point imaging ;

[0046] Finally .

[0047] The identity analysis method based on human gait video has the beneficial effects that the application first establishes a space reference coordinate system through four groups of cameras, and instructs personnel to walk along the route, so that the four groups of cameras can record the front, back, left and right of the personnel, and mark the key points of the personnel in the walking process, so that multiple marks of a single key point can be obtained after the personnel finish walking, the displacement amount of unit time is calculated along the advancing direction of the route through multiple marks of the key points, so that the average speed is calculated, when the unit time is small enough, the average speed can be approximated as the speed of the midpoint of the displacement, and the instantaneous speed of the human key point at each place along the advancing direction of the route can be calculated by analogy, since the time difference between adjacent instantaneous speeds is also the unit time, the acceleration of the human key point at each place along the advancing direction of the route is equal to the ratio between the difference of instantaneous speeds and the time difference, so that the horizontal acceleration and the vertical acceleration can be obtained, the horizontal acceleration and the vertical acceleration of the key point are taken as gait features, after the gait features are bound with the personnel identity, the gait features are input into the database, and the human gait feature database is established. The gait features of the personnel are collected by establishing the human gait feature database, so that the gait features of the target personnel can be compared conveniently in the later stage.

[0048] In order to identify the target personnel, therefore, any one segment of the monitoring video of the target area is acquired, and the parameters of the camera corresponding to the monitoring video are acquired, and the coordinate system of the camera is established; by marking the key nodes of each step gait of the target personnel in the monitoring video, and calculating the coordinates of the key nodes in the coordinate system in each step gait of the target personnel, and then calculating the horizontal acceleration and the vertical acceleration of the key nodes in the coordinate system when walking continuously, the gait features of the human body are obtained by solving the acceleration of the key nodes in each coherent action of the target personnel when walking continuously. Since the motion of each key node of each person when walking is not the same, the gait features of each person can be identified as unique gait features of each person, so that similar target personnel can be compared in the pre-established human gait feature database, and the identity of the target personnel in the monitoring video is further confirmed by the human eye recognition of the staff, the identity of the target personnel is confirmed, and the personnel in close contact with the target personnel is determined, especially the identity of the target personnel wearing a mask is confirmed. Since it is the identification of gait features, even if a mask is worn, the gait features will not change, so that the trajectory is quickly traced, the social work is ensured to be normal and safe, and the social work is ensured to be normal and safe. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 For the identity analysis method of the human gait video based on the present application;

[0050] Figure 2 For the identity analysis method of the human gait video based on the present application;

[0051] Figure 3 For the device layout of the human gait feature database established in the embodiment of the present application;

[0052] Figure 4 For the road plane coordinate system of the embodiment of the present application;

[0053] Figure 5 For Figure 4 The image plane coordinate system corresponding to the road plane of the embodiment;

[0054] Figure 6 For the imaging model of the Y-axis direction of the coordinate system of the embodiment of the present application;

[0055] Figure 7 For the imaging model of the X-axis direction of the coordinate system of the embodiment of the present application. DETAILED DESCRIPTION

[0056] The present application will be further described below in conjunction with the drawings and specific embodiments.

[0057] Reference Figures 1 to 3As shown, the present application provides a human gait video-based identity analysis method, comprising the following steps:

[0058] Step one, establish a human gait feature database, the specific steps are as follows:

[0059] Step 1.1. Set up a straight route, and set up four groups of cameras to establish a space reference coordinate system, wherein two groups of cameras are symmetrically arranged on both sides of the route, and the camera optical axes of the cameras are perpendicular to the forward direction of the route, and the other two groups of cameras are arranged at the starting point and the ending point of the route, and the camera optical axes are parallel to the route, and the space reference coordinate system takes the starting point of the route as the origin, and takes the direction along the route, the direction perpendicular to the route and the direction perpendicular to the route as the space rectangular coordinate axes;

[0060] Step 1.2. Instruct personnel to walk straight from the starting point to the ending point on the route, and then mark the joint nodes of each step of the personnel gait;

[0061] Step 1.3. According to the image data of the four groups of cameras, the coordinates of the marked joint nodes in the space reference coordinate system are calculated, and then the horizontal acceleration and vertical acceleration of the joint nodes in the space reference coordinate system in each step of the continuous walking of the personnel are calculated;

[0062] Step 1.4. The horizontal acceleration and vertical acceleration of the joint nodes in the space reference coordinate system are taken as the gait features, and the gait features are bound with the personnel identity, and then entered into the database to establish a human gait feature database;

[0063] Step two, process the video to be identified, the specific steps are as follows:

[0064] Step 2.1. Obtain any one section of the monitoring video in the target area, and obtain the parameters of the corresponding camera of the monitoring video, and establish the coordinate system of the camera;

[0065] Step 2.2. Mark the joint nodes of each step of the target personnel in the monitoring video obtained in step 2.1, and calculate the coordinates of the joint nodes in the coordinate system in each step of the target personnel, and then calculate the horizontal acceleration and vertical acceleration of the joint nodes in the coordinate system in continuous walking;

[0066] Step 2.3. Compare the horizontal acceleration and vertical acceleration of the joint nodes obtained in step 2.2 in the coordinate system with the human gait feature database, and rank according to the similarity, and specially mark the target personnel with higher similarity;

[0067] Step three, identify the identity of the target personnel specially marked in step two by artificial.

[0068] Since the human gait feature database can obtain the required parameter data when being established, the condition that the coordinates of the human joint points in each frame of the image data are located in the established space reference coordinate system is met, so that the space coordinates of each joint point of the human body in the continuous walking action can be solved by a large amount of labeling of the image data, and the vector acceleration of each joint point in the human motion can be solved by relying on the space displacement of each joint point in the continuous action and the required time, so that the characteristic motion state of each joint point can be drawn.

[0069] Specifically in the embodiment, the space reference coordinate system is first established by the four groups of cameras, and the personnel are instructed to walk along the route, so that the four groups of cameras can record the front, back, left and right of the personnel, and label the joint points of the personnel in the walking process, so that multiple labels of a single joint point can be obtained after the personnel finish walking, the displacement in unit time is calculated along the advancing direction of the route by the multiple labels of the joint points, so that the average speed is solved, and when the unit time is small enough, the average speed can be approximated as the speed of the midpoint of the displacement, and the instantaneous speed of the human joint point at each place along the advancing direction of the route can be solved by analogy. Since the time difference between adjacent instantaneous speeds is also the unit time, the acceleration of the human joint point at each place along the advancing direction of the route is equal to the ratio between the difference of the instantaneous speeds and the time difference, so that the horizontal acceleration and the vertical acceleration can be obtained. The horizontal acceleration and the vertical acceleration of the joint point are taken as the gait features, the gait features are bound with the identity of the personnel, and are input into the database to establish the human gait feature database. The gait features of the personnel are collected by establishing the human gait feature database, so that the gait features of the target personnel can be compared conveniently in the later stage.

[0070] In order to identify the target personnel, therefore, any one segment of the monitoring video of the target area is acquired, and the parameters of the camera corresponding to the monitoring video are acquired, and the coordinate system of the camera is established; by marking the key nodes of each step gait of the target personnel in the monitoring video, and calculating the coordinates of the key nodes in the coordinate system in each step gait of the target personnel, and then calculating the horizontal acceleration and vertical acceleration of the key nodes in the coordinate system when walking continuously; by solving the acceleration of the key nodes in each coherent action of the target personnel when walking continuously in the monitoring video, the gait features of the human body are obtained, since the motion of each key node of each person when walking is not the same, so it can be identified as the gait feature unique to each person, so that similar target personnel can be compared in the pre-established human gait feature database, and further confirmed by the human eye recognition of the staff, assisting the staff to confirm the identity of the target personnel in the monitoring video, determining the target personnel, and determining the personnel in close contact with the target personnel, especially assisting in confirming the identity of the target personnel wearing a mask, because it is the identification of gait features, so even if wearing a mask, the gait features will not change, thereby quickly tracing, determining the trajectory, and ensuring that social work is normal and safe.

[0071] Specifically, in step one, the height, gender, body shape, and personal habitual action of the personnel are bound to the personnel identity. By binding the height, gender, body shape, and personal habitual action of the personnel to the personnel identity, the features of the personnel identity are increased, and the above-mentioned features are saved to the established human gait feature database. When the staff further identifies by the human eye, the above-mentioned features can be used as a judgment basis for identification, thereby increasing the accuracy of identification.

[0072] Specifically, in step three, the specific steps of artificial identity recognition are as follows:

[0073] Step 3.1. Identify the gender of the specially marked target personnel and exclude those who do not meet the gender requirements.

[0074] Step 3.2. Identify the height of the specially marked target personnel and exclude those who do not meet the height requirements.

[0075] Step 3.3. Identify the body shape of the specially marked target personnel and exclude those who do not meet the body shape requirements.

[0076] Step 3.4. Identify the personal habitual action of the specially marked target personnel and confirm the target personnel identity in combination with the personal habitual action.

[0077] Since the height, gender, body shape, and personal habitual action of the person are bound to the identity of the person in step one, the above features are saved in the human gait feature database. The staff can identify the specially marked target person through the height, gender, body shape, and personal habitual action of the person in the human gait feature database. The height, gender, body shape, and personal habitual action are all relatively obvious features, so the staff can directly judge by naked eye. Through the judgment of the above personnel identity features, the specially marked target person is further screened, and the recognition accuracy is improved.

[0078] Specifically, in step two, the obtained monitoring video should meet the condition of containing at least 5 continuous normal walking steps of the target person.

[0079] Specifically, in step two, the joint nodes of the target person in the monitoring video are marked at a frame rate period of 15-30 FPS. Since a monitoring video is composed of multiple continuous images, the video of each step of the target person is composed of multiple continuous images. In order to analyze the gait features of each step of the target person, the joint nodes of the target person in the monitoring video are marked at a frame rate period of 15-30 FPS, so that multiple accelerations of the joint nodes in different positions of the target person in one step can be obtained, and multiple accelerations can be extracted, that is, more gait feature data can be obtained. The increase of gait feature data increases the data used for identifying the target person, so that the gait features and the target person are more accurate, and the recognition accuracy is improved.

[0080] Specifically, the joint nodes include wrist joint, elbow joint, shoulder joint, neck joint, ankle joint, knee joint, and hip joint. By marking the above joint nodes, the joint nodes involved in the movement of the person during walking can be marked, so that the complete gait features of the human body during walking can be obtained, and the recognition accuracy is improved.

[0081] Reference Figures 4 to 7As shown in step two, a coordinate system is established, the road plane is a triangular plane ABU, ABCD is a trapezoidal area on the road plane shot by the camera, CD intersects with BU at point C, CD intersects with AU at point D, CD is parallel to AB, O is the center point of the camera lens, OG is the camera optical axis, G is the intersection point of the camera optical axis and the road plane, which is also the intersection point of the diagonal of the field of view trapezoid ABCD, I is the vertical projection of O on the road plane, the height of the camera OI is h, UE intersects with AB at point E and is perpendicular to AB, UE intersects with CD at point F, JK intersects with UE at point G, JK intersects with AU at point K, JK intersects with BU at point J, JK is parallel to AB, in the road plane coordinate system, G is defined as the origin of the coordinate system, UE is defined as the Y axis of the coordinate system, the forward direction of the person is defined as the Y axis direction, and JK is defined as the X axis of the coordinate system;

[0082] The G, A, B, C, D points of the road plane correspond to g, a, b, c, d points on the image plane respectively, a, b, c, d are the four end points of the rectangle, g is the midpoint of the image plane, ef intersects with ab at e point, ef intersects with cd at f point, jk intersects with ef at g point, jk intersects with bc at j point, jk intersects with ad at k point, the E, F, J, K points of the road plane correspond to e, f, j, k points on the image plane respectively, the height of the image plane cb is H, the width of the image plane cd is W, the g point of the image plane is defined as the origin of the image plane coordinate system, ef is the y axis of the image plane coordinate system, the forward direction of the person is the y axis direction, and jk is the x axis of the image plane coordinate system, taking a point P on the road plane, the coordinates of P point in the road plane coordinate system are (X p , Y p ), P point corresponds to p point on the image plane, the coordinates of p point in the image plane coordinate system are (x p , y p );

[0083] The plane perpendicular to the camera optical axis and the road plane is OEI, ML intersects with OG at G point and is perpendicular, ML intersects with the extension of fF at L point, py represents the projection of p point on the image plane on ef, P y point is the projection of P point on the road plane on the Y axis, Z point is the intersection of p y P y and ML, the horizontal field of view angle of the camera 2∠BUE is 2β0, the vertical field of view angle 2∠EOG is 2α0, the pitch angle ∠GOI is γ0, ∠P y OG is α, and ∠P y OP is α1;

[0084] The imaging model in the Y axis direction can be obtained according to the trigonometric function relationship and similar triangles, ;

[0085] Further ;

[0086] Further ;

[0087] X-axis direction imaging model, P x is the projection point of P point on X axis, in triangle △OP y L, ;

[0088] According to the similar triangle imaged by P point ;

[0089] Finally .

[0090] The embodiment of the application also provides an identity analysis method based on human gait video, comprising the following steps:

[0091] Step one, since the movements of the joints of a person are not the same when walking, the movements of the joints are identified as the gait features unique to each person, thereby establishing a human gait feature database;

[0092] Step two, any monitoring video of a target area is obtained, and the horizontal acceleration and the vertical acceleration of each joint of the target person in each gait are calculated in the camera coordinate system corresponding to the monitoring video, and are used for comparison with the human gait feature database, and the first few target persons with higher similarity are specially marked;

[0093] Step three, manual exclusion confirms the identity, by solving the acceleration of each joint when continuously walking, the acceleration can be identified as the gait features unique to each person, thereby assisting the staff to confirm the identity of the target person in the monitoring video, determining the confirmed person, and thereby determining the close contact person.

[0094] Specifically, in the step two, a coordinate system is established, the road plane is a triangular plane ABU, ABCD is a trapezoidal area on the road plane shot by the camera, CD intersects with BU at C point, CD intersects with AU at D point, CD is parallel to AB, O point is the lens center point of the camera, OG is the optical axis of the camera, G point is the intersection point of the optical axis of the camera and the road plane, and is also the intersection point of the diagonal of the field of view trapezoid ABCD, I point is the vertical projection of O point on the road plane, the height OI of the camera is h, UE intersects with AB at E point, and UE is perpendicular to AB, UE intersects with CD at F point, JK intersects with UE at G point, JK intersects with AU at K point, JK intersects with BU at J point, JK is parallel to AB, in the road plane coordinate system, G point is defined as the origin of the coordinate system, UE is defined as the Y axis of the coordinate system, the forward direction of the person is defined as the Y axis direction, and JK is defined as the X axis of the coordinate system;

[0095] G, A, B, C, D points on the road plane correspond to g, a, b, c, d points on the image plane respectively, a, b, c, d are four end points of the rectangle, g is the midpoint of the image plane, ef intersects ab at e point, ef intersects cd at f point, jk intersects ef at g point, jk intersects bc at j point, jk intersects ad at k point, E, F, J, K points on the road plane correspond to e, f, j, k points on the image plane respectively, the height of the image plane cb is H, the width of the image plane cd is W, define the g point on the image plane as the coordinate origin of the image plane coordinate system, ef is the y axis of the image plane coordinate system, the forward direction of the personnel is the y axis direction, jk is the x axis of the image plane coordinate system, take a point P on the road plane, the coordinates of P point on the road plane coordinate system are (X p , Y p ), P point corresponds to p point on the image plane, the coordinates of p point on the image plane coordinate system are (x p , y p );

[0096] The plane perpendicular to the road plane of the camera optical axis is OEI, ML intersects OG at G point and is perpendicular, ML intersects the extension of fF at L point, p y represents the projection of p point on the image plane on ef, P y point is the projection of P point on the road plane on Y axis, Z point is the intersection of p y P y and ML, the horizontal field of view angle of the camera 2 ∠BUE is 2β0, the vertical field of view angle 2 ∠EOG is 2α0, the pitch angle ∠GOI is γ0, ∠PyOG is α, ∠P y OP is α1;

[0097] The imaging model in Y axis direction can be obtained according to the trigonometric function relationship and similar triangles, ;

[0098] Further ;

[0099] Further ;

[0100] The imaging model in X axis direction, P x is the projection point of P point on X axis, in triangle △OP y L, ;

[0101] According to the similar triangle of P point imaging, ;

[0102] Finally .

[0103] The road plane is a plane where the target person actually locates, the image plane is a plane where the target person locates in the monitoring video, and the like plane is a mapping of the road plane in the camera, so that the actual gait feature of the target person can be obtained through the monitoring video of the camera, and the above formula can be obtained through the mapping relationship between the road plane and the image plane and the parameters of the camera, and the position of the joint of the target person in the walking process can be obtained through the above formula, and the acceleration of the joint can be obtained through the movement of the joint position.

[0104] The above is only a preferred embodiment of the present application, and does not limit the technical scope of the present application in any way. Any slight modification, equivalent change and modification of the above embodiment according to the technical essence of the present application still falls within the scope of the technical solution of the present application.

Claims

1. An identity analysis method based on human gait video, characterized in that, Includes the following steps: Step 1: Establish a human gait feature database; Step 2: Process the video to be identified. The specific steps are as follows: Step 2.

1. Obtain any segment of surveillance video in the target area, and obtain the parameters of the camera corresponding to the surveillance video, and establish the coordinate system of the camera; Step 2.

2. Mark the joints of each step of the target person's gait in the surveillance video obtained in Step 2.1, and calculate the coordinates of the joints in the coordinate system for each step of the target person's gait. Then calculate the horizontal acceleration and vertical acceleration of the joints in the coordinate system when walking continuously. Step 2.

3. Compare the horizontal and vertical accelerations of the joint points obtained in Step 2.2 in the coordinate system with the human gait feature database, rank them according to similarity, and specially mark the target personnel with high similarity. In step two, a coordinate system is established. The road plane is a triangular plane ABU. ABCD is a trapezoidal region on the road plane captured by the camera. CD intersects BU at point C, CD intersects AU at point D, and CD is parallel to AB. Point O is the center point of the camera lens. OG is the optical axis of the camera. Point G is the intersection of the camera's optical axis and the road plane, and also the intersection of the diagonals of the field-of-view trapezoid ABCD. Point I is the vertical projection of point O onto the road plane. The height of the camera OI is h. UE intersects AB at point E, and UE is perpendicular to AB. UE intersects CD at point F. JK intersects UE at point G, JK intersects AU at point K, JK intersects BU at point J, and JK is parallel to AB. In the road plane coordinate system, point G is defined as the origin, UE is defined as the Y-axis, the direction of personnel movement is defined as the Y-axis direction, and JK is defined as the X-axis. Points G, A, B, C, and D on the road plane correspond to points g, a, b, c, and d respectively on the image plane of the rectangle. a, b, c, and d are the four endpoints of the rectangle, and g is the midpoint of the image plane. ef intersects ab at point e, ef intersects cd at point f, jk intersects ef at point g, jk intersects bc at point j, and jk intersects ad at point k. Points E, F, J, and K on the road plane correspond to points e, f, j, and k respectively on the image plane. The height cb of the image plane is H, and the width cd of the image plane is W. Point g on the image plane is defined as the origin of the image plane coordinate system, ef is the y-axis of the image plane coordinate system, the direction of movement of personnel is the y-axis direction, and jk is the x-axis of the image plane coordinate system. A point P is taken on the road plane, and the coordinates of point P in the road plane coordinate system are (X...). p Y p Point P corresponds to point p in the image plane, and the coordinates of point p in the image plane coordinate system are (x, y). p y p ); The plane perpendicular to the road plane and the optical axis of the camera is OEI. ML and OG intersect at point G and are perpendicular. ML intersects the extension of fF at point L. y Let P represent the projection of point p on the image plane onto ef. y Point P is the projection of point P on the road plane onto the Y-axis, and point Z is point P. y P y At the intersection with ML, the camera's horizontal field of view 2∠BUE is 2β0, the vertical field of view 2∠EOG is 2α0, the pitch angle ∠GOI is γ0, and ∠P y OG is α, ∠P y OP is α1; The imaging model in the Y-axis direction can be obtained based on trigonometric relationships and similar triangles. ; and then ; and then ; Imaging model in the X-axis direction, P x Let P be the projection of point P onto the X-axis, within triangle △OP. y In L, ; Based on the similar triangles formed by the images of point P, we can obtain... ; at last ; Step 3: The target personnel specifically marked in Step 2 are identified manually.

2. The identity analysis method based on human gait video according to claim 1, characterized in that, In step one, a human gait feature database is established. The specific steps are as follows: Step 1.

1. Establish a straight travel route and set up four sets of cameras to establish a spatial reference coordinate system. Two sets of cameras are symmetrically arranged on both sides of the travel route, and the optical axes of the cameras are perpendicular to the travel direction. The other two sets of cameras are respectively set at the start and end points of the travel route, and the optical axes of the cameras are parallel to the travel route. The spatial reference coordinate system takes the start point of the travel route as the origin, and uses the direction along the travel route, the horizontal direction perpendicular to the travel route, and the vertical direction perpendicular to the travel route as spatial rectangular coordinate axes. Step 1.

2. Instruct the personnel to walk straight from the starting point to the end point along the route, and then mark the key points of each step of the personnel's gait; Step 1.

3. Calculate the coordinates of the marked joint points in the spatial reference coordinate system based on the image data of the four sets of cameras, and then calculate the horizontal and vertical accelerations of the joint points in the spatial reference coordinate system in each step of the person's continuous walking gait. Step 1.

4. The horizontal and vertical accelerations of the joint points in the spatial reference coordinate system are used as gait features. After binding the gait features with the person's identity, they are entered into the database to establish the human gait feature database.

3. The identity analysis method based on human gait video according to claim 2, characterized in that, Step one also includes binding the person's height, gender, body shape, and personal habitual movements with the person's identity.

4. The identity analysis method based on human gait video according to claim 1, characterized in that, In step three, the specific steps of manual identity verification are as follows: Step 3.

1. Identify the gender of the specially marked target personnel and exclude those whose gender does not match; Step 3.

2. Identify the height of the specially marked target personnel and exclude those whose height does not match; Step 3.

3. Identify the body posture of the specially marked target personnel and exclude those whose body posture does not conform; Step 3.

4. Identify the personal habitual actions of the specially marked target personnel, and confirm the identity of the target personnel by combining the personal habitual actions.

5. The identity analysis method based on human gait video according to claim 1, characterized in that, In step two, the acquired surveillance video should meet the condition that it shows the target person walking continuously and normally for at least 5 steps.

6. The identity analysis method based on human gait video according to claim 1, characterized in that, In step two, the key points of the target person in the surveillance video are marked with a frame rate period of 15-30 FPS.

7. The identity analysis method based on human gait video according to claim 1, characterized in that, The joints include the wrist, elbow, shoulder, neck, ankle, knee, and hip joints.

8. An identity analysis method based on human gait video, characterized in that, Includes the following steps: Step 1: Since the movement of each joint point is different when a person walks, the movement of each joint point is identified as a unique gait feature of each person, thereby establishing a human gait feature database. Step 2: Obtain any segment of surveillance video in the target area, and calculate the horizontal and vertical accelerations of each joint point in the target person's gait in the camera coordinate system corresponding to the surveillance video. Compare these accelerations with the human gait feature database, and mark the top few target persons with the highest similarity. In step two, a coordinate system is established. The road plane is a triangular plane ABU. ABCD is the trapezoidal region on the road plane captured by the camera. CD intersects BU at point C, CD intersects AU at point D, and CD is parallel to AB. Point O is the center point of the camera lens. OG is the optical axis of the camera. Point G is the intersection of the camera's optical axis and the road plane, and also the intersection of the diagonals of the field-of-view trapezoid ABCD. Point I is the vertical projection of point O onto the road plane. The height of the camera OI is h. UE intersects AB at point E, and UE is perpendicular to AB. UE intersects CD at point F. JK intersects UE at point G, JK intersects AU at point K, JK intersects BU at point J, and JK is parallel to AB. In the road plane coordinate system, point G is defined as the origin of the coordinate system, UE is defined as the Y-axis of the coordinate system, the direction of personnel movement is defined as the Y-axis direction, and JK is defined as the X-axis of the coordinate system. Points G, A, B, C, and D on the road plane correspond to points g, a, b, c, and d respectively on the image plane of the rectangle. a, b, c, and d are the four endpoints of the rectangle, and g is the midpoint of the image plane. ef intersects ab at point e, ef intersects cd at point f, jk intersects ef at point g, jk intersects bc at point j, and jk intersects ad at point k. Points E, F, J, and K on the road plane correspond to points e, f, j, and k respectively on the image plane. The height cb of the image plane is H, and the width cd of the image plane is W. Point g on the image plane is defined as the origin of the image plane coordinate system, ef is the y-axis of the image plane coordinate system, the direction of movement of personnel is the y-axis direction, and jk is the x-axis of the image plane coordinate system. A point P is taken on the road plane, and the coordinates of point P in the road plane coordinate system are (X...). p Y p Point P corresponds to point p in the image plane, and the coordinates of point p in the image plane coordinate system are (x, y). p y p ); The plane perpendicular to the road plane and the optical axis of the camera is OEI. ML and OG intersect at point G and are perpendicular. ML intersects the extension of fF at point L. y Let P represent the projection of point p on the image plane onto ef. y Point P is the projection of point P on the road plane onto the Y-axis, and point Z is point P. y P y At the intersection with ML, the camera's horizontal field of view 2∠BUE is 2β0, the vertical field of view 2∠EOG is 2α0, the pitch angle ∠GOI is γ0, and ∠P y OG is α, ∠P y OP is α1; The imaging model in the Y-axis direction can be obtained based on trigonometric relationships and similar triangles. ; and then ; and then ; Imaging model in the X-axis direction, P x Let P be the projection of point P onto the X-axis, within triangle △OP. y In L, ; Based on the similar triangles formed by the images of point P, we can obtain... ; at last ; Step 3: Manual identification and screening. By calculating the acceleration of each joint point during continuous walking, it can be identified as a unique gait characteristic of each person, thereby assisting staff in confirming the identity of the target person in the surveillance video, identifying confirmed cases, and thus identifying close contacts.

Images