Abnormal event generation method and device, equipment and medium

Abstract

The invention discloses an abnormal event generation method and device, equipment and a medium, and the method comprises the steps: obtaining a target image, carrying out the limb analysis of each human body object in the target image, and obtaining the limb feature data of each human body object; detecting whether the limb feature data of each human body object in the at least one human body object meets an abnormal action condition, and generating an abnormal event when it is detected that the limb feature data in the limb feature data of all the human body objects meets the abnormal action condition, according to the invention, real-time analysis and discovery of abnormal conditions such as fighting in the monitoring video are automatically realized, the accuracy of monitoring the abnormal conditions is improved, and occurrence of public safety hazards is prevented from the source.

Description

Technical Field

[0001] The present invention relates to the field of intelligent video surveillance, and in particular to an abnormal event generation method, device, equipment and medium. Background Art

[0002] With the development of video surveillance, more and more surveillance cameras are used in public places to ensure public safety. Traditional video surveillance systems mainly rely on several staff members to monitor multiple surveillance devices at the same time to monitor whether there are abnormal situations such as fighting in the surveillance images in real time. However, with the increase in the amount of surveillance video data, monitoring surveillance video data usually requires a lot of time and manpower costs. In addition, due to the complexity of the surveillance image content and the limited personal energy of the monitoring personnel, there may be omissions in monitoring abnormal situations or untimely detection, which may lead to the occurrence of incidents that endanger public safety. Therefore, how to automatically and accurately detect abnormal situations in surveillance videos in real time, so as to deal with them as soon as possible and ensure the safety of public places has become an urgent problem to be solved. Summary of the invention

[0003] Based on this, a method, device, equipment and medium for generating abnormal events are provided to solve the problem of how to automatically, accurately and in real time discover abnormal situations in surveillance videos, so as to carry out corresponding processing as soon as possible according to the abnormal situations and ensure the safety of public places.

[0004] In a first aspect, an embodiment of the present invention provides a method for generating an abnormal event, the method comprising the following steps:

[0005] Acquire a target image, where the target image is an image that meets a crowd gathering condition and includes at least one human object;

[0006] Performing limb analysis on each human object in the target image to obtain limb feature data of each human object;

[0007] Detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition;

[0008] When it is detected that among the limb feature data of all human objects, limb feature data satisfies the abnormal action condition, an abnormal event is generated.

[0009] In a second aspect, an embodiment of the present invention provides an abnormal event generating device, the abnormal event generating device comprising:

[0010] A first acquisition module is used to acquire a target image, wherein the target image is an image that meets a crowd gathering condition and includes at least one human object;

[0011] An analysis module, used for performing limb analysis on each human object in the target image to obtain limb feature data of each human object;

[0012] A first detection module, used to detect whether the limb feature data of each human object in the at least one human object meets the abnormal action condition;

[0013] The generating module is used to generate an abnormal event when it is detected that the limb feature data of all human objects meets the abnormal action condition.

[0014] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned abnormal event generation method when executing the computer program.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned abnormal event generation method are implemented.

[0016] The technical effect achieved by the present invention is different from that of the prior art: the present invention obtains a target image that meets the crowd gathering condition, performs limb analysis on each human object in the target image, detects whether the limb feature data of each human object in at least one human object meets the abnormal action condition, and generates an abnormal event when it is detected that there is limb feature data that meets the abnormal action condition in the limb feature data of all human objects. Among them, by detecting and analyzing the limb feature data of each human object in the target image, comprehensive monitoring of each human object in the monitoring screen is achieved, and the accuracy of monitoring and detecting abnormal situations such as fighting is improved. At the same time, by generating corresponding abnormal events based on the detection and analysis results of the limb feature data of each human object, relevant personnel can be reminded early to handle the abnormal event accordingly, thereby preventing the occurrence of public safety incidents from the root. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0018] Figure 1 This is a schematic diagram of an application environment of an abnormal event generation method provided by Embodiment 1 of the present invention;

[0019] Figure 2 It is a flowchart of a method for generating an abnormal event provided in Embodiment 1 of the present invention;

[0020] Figure 3 It is a flowchart of a method for generating an abnormal event provided in Embodiment 2 of the present invention;

[0021] Figure 4 It is a flowchart of a method for generating an abnormal event provided by Embodiment 3 of the present invention;

[0022] Figure 5 It is a flowchart of a method for generating an abnormal event provided by Embodiment 4 of the present invention;

[0023] Figure 6 It is a flowchart of a method for generating an abnormal event provided in Embodiment 5 of the present invention;

[0024] Figure 7 It is a flowchart of a method for generating an abnormal event provided by Embodiment 6 of the present invention;

[0025] Figure 8 1 is a flow chart of a method for generating an abnormal event provided in Embodiment 7 of the present invention;

[0026] Fig. 9 It is a flowchart of a method for generating an abnormal event provided by Embodiment 8 of the present invention;

[0027] Fig.10 It is a structural schematic diagram of an abnormal event generating device provided by Embodiment 9 of the present invention;

[0028] Fig.11 It is a structural diagram of a computer device provided in Embodiment 10 of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] It should be understood that the order of execution of the steps in the following embodiments does not imply a precedence of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0031] See also Figure 1, is a schematic diagram of an application environment of an abnormal event generation method provided in the first embodiment of the present invention, wherein the abnormal event generation method is to analyze image or video data to obtain corresponding analysis results, wherein the client communicates with the server. The client includes but is not limited to a palmtop computer, a desktop computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a server computer device, a personal digital assistant (PDA) and other devices, and the client can also be an image acquisition device such as a camera or a still camera. The server can be implemented by an independent server or a server cluster composed of multiple servers, and can be specifically embodied as a computer device.

[0032] In a possible implementation, the client may capture an image and then process the image to identify an abnormal event and generate an abnormal event.

[0033] See also Figure 2 , is a flow chart of an abnormal event generation method provided in Embodiment 1 of the present invention, and the abnormal event generation method can be applied in Figure 1 In the application environment shown, the server obtains the image or video (i.e., continuous image) data sent by the client to analyze the image or video data. Of course, the client can be an image acquisition device, or a mobile phone, computer, etc. These devices are connected to the corresponding image acquisition device to obtain the image to be processed. In addition, the server can directly obtain the corresponding image or video from its database. Of course, this part of the image or video is also sent by the client or other servers.

[0034] like Figure 2 As shown, the abnormal event generation method includes the following steps:

[0035] Step S201: Acquire a target image, where the target image is an image that meets a crowd gathering condition and includes at least one human object.

[0036] In this embodiment, the server obtains video data, and by screening each frame of the obtained video data, obtains an image in which the number of people reaches a preset gathering number threshold or the density of people in a certain area reaches a preset gathering density threshold, and by identifying all or any unique features of the human body in the image, the number of people in the image can be determined. Among them, reaching the gathering number threshold or density threshold is to meet the above-mentioned crowd gathering conditions. Of course, the crowd gathering conditions can also be set according to other requirements.

[0037] A human object may refer to an object characterized as a human body. After obtaining a target image, each human object may be obtained through a target detection algorithm, and each human object may be labeled or identified so that each human object can be effectively distinguished during subsequent analysis.

[0038] For example, the server connects to the corresponding video acquisition device to obtain real-time video data, performs preprocessing operations such as denoising and image enhancement on each frame of the acquired video data, and then determines the number of people in each image based on facial features. For example, the number of people in image 1 is 8, and the number of people is compared with the aggregation number threshold (for example, the aggregation number threshold is 5). It can be seen that Figure 1 The number of people exceeds the gathering threshold, so the image 1 is determined to be the target image. Figure 1 In particular, a target detection algorithm is used to obtain each human object, including assigning a label to each human object, and mapping the label with the regional position information of the corresponding human object in image 1, etc., to facilitate subsequent analysis.

[0039] Step S202: Perform limb analysis on each human object in the target image to obtain limb feature data of each human object.

[0040] In this embodiment, the limb feature data may be data related to the limb parts and features of the human object, wherein the analysis of the human object may be divided into the following steps: 1) using image segmentation and classification to determine the limbs (marking the left and right), head, and torso; 2) extracting the image features corresponding to each limb, specifically extracting the pixel position and pixel value of the corresponding limb in the image; 3) obtaining feature data that can be directly expressed based on the extracted features, for example, determining the upper arm length and forearm length of the left arm, determining the facial feature data and head width of the head, etc.; of course, the position of the limbs may also be recorded to facilitate the calculation of the positional relationship between the limbs.

[0041] For example, for the left upper limb, based on the directly extracted length of the left upper limb, combined with the triangle cosine theorem and the speed calculation formula, data related to the left upper limb characteristics such as the elbow bending angle of the left upper limb and the swinging speed of the left upper limb can be calculated.

[0042] Step S203: Detect whether the limb feature data of each human object in at least one human object meets the abnormal action condition.

[0043] Step S204: when it is detected that there is limb feature data satisfying the abnormal action condition among the limb feature data of all human objects, an abnormal event is generated.

[0044] In this embodiment, the abnormal action condition may be that the height of the leg is greater than the height threshold of the leg, the angle of the legs is within the angle range of the legs, and the speed of the arms is greater than the speed threshold of the arms. When the limb data meets at least one of the abnormal action conditions, an abnormal event exists, and the abnormal event may refer to fighting, theft, and gambling in a group, etc. Among them, each abnormal action may correspond to one or more abnormal events. For example, if the abnormal action is kicking, the abnormal event corresponding to the kicking action may be a single abnormal event of fighting, or two abnormal events of fighting and gambling in a group.

[0045] For example, if the limb feature data is the speed of arm swinging, the abnormal action condition is that the arm swinging speed is greater than the arm swinging speed threshold. For example, the preset arm swinging speed threshold is 2 times per second, and the detected arm swinging speed is 3 times per second. When the arm swinging speed meets the abnormal action condition, an abnormal event is generated.

[0046] In this embodiment, by acquiring a target image that meets the crowd gathering condition, performing limb analysis on each human object in the target image, detecting whether the limb feature data of each human object in at least one human object meets the abnormal action condition, and generating an abnormal event when it is detected that there is limb feature data that meets the abnormal action condition in the limb feature data of all human objects. Among them, by detecting and analyzing the limb feature data of each human object in the target image, comprehensive monitoring of each human object in the monitoring screen is achieved, and the accuracy of monitoring and detecting abnormal situations such as fighting is improved. At the same time, by generating corresponding abnormal events based on the detection and analysis results of the limb feature data of each human object, relevant personnel can be reminded to handle the abnormal event accordingly as early as possible, thereby preventing the occurrence of public safety incidents from the root.

[0047] The abnormal event generation method of the above-mentioned embodiment is applied to an application environment in which a server interacts with a client. Of course, in one embodiment, the abnormal time generation method can be directly applied to an image acquisition device. For example, the image acquisition device is a camera with an implanted algorithm chip. The camera can directly identify abnormal events on the acquired images and then upload the identification results to the abnormal event management platform. This process does not require the acquired images to be transmitted to the server for identification, but can directly give the identification results.

[0048] See also Figure 3 , which is a flow chart of an abnormal event generation method provided in Embodiment 2 of the present invention, wherein the step S203 in Embodiment 1 detects whether the limb feature data of each human object in at least one human object meets the abnormal action condition, and may include the following steps:

[0049] Step S301: for any human object, when detecting that the limb feature data of the human object meets the leg lifting feature, an adjacent image of the target image is obtained.

[0050] In this embodiment, the leg lifting feature may refer to a leg lifting action in which the foot of the corresponding leg leaves the ground. In the feature extraction process, the position of the foot in the image and the position of the ground are obtained, and based on the positional relationship between the two, it can be determined whether the foot is in contact with the ground. In this process, the limb feature data required may include data such as the position feature of the foot and the position feature of the ground.

[0051] For the target image, its adjacent image may refer to an image captured at a similar acquisition time to the target image. If any human object in the target image has a leg-lifting action, it is necessary to find the moment of leg-lifting. Therefore, it is necessary to obtain the image captured before the acquisition time of the target image. Specifically, according to the preset time period, the image captured within the preset time period before the acquisition time of the target image is obtained, which is the adjacent image before the target image.

[0052] If there is a leg-lifting action in any human body image in the target image, it is necessary to find the end moment of the leg-lifting action. Therefore, it is necessary to obtain the image captured after the acquisition time of the target image. Specifically, according to the preset time period, the image captured within the preset time period after the acquisition time of the target image is acquired, which is the adjacent image after the target image.

[0053] For example, human object A in image 1 is detected, and according to the limb feature data of human object A, it is determined that the height of the feet of human object A from the ground is 0.02 meters, and the height is greater than 0.01 meters, and it is determined that the feet of human object A are off the ground. The images captured within 10 minutes before the acquisition time of image 1 are obtained, which are the adjacent images before image 1, and the images captured within 10 minutes after the acquisition time of image 1 are obtained, which are the adjacent images after image 1.

[0054] Step S302: Obtain the lifting start time corresponding to the raised leg of the human subject according to the acquisition time of all adjacent images and the acquisition time of the target image.

[0055] In this embodiment, the acquisition time may be the acquisition time corresponding to each adjacent image at the video acquisition device. Each adjacent image is identified to determine the position of the foot of the corresponding human object in the image and the position of the ground. According to the positional relationship between the two, the distance between the foot and the ground can be determined. If the distance is greater than the preset distance from the ground, it is determined that the foot is off the ground, and the acquisition time of the previous image of the first off-the-ground image is the lifting start time.

[0056] For example, the preset distance from the ground is 0.01 meters, and a distance greater than 0.01 meters can be determined as the feet are off the ground, and the acquisition time of the image before the first off-the-ground image is the lifting start time. If the distance between the feet of the corresponding human object and the ground in the adjacent images before the target image is 0, the distance between the feet of the corresponding human object and the ground in the target image is less than 0.01 meters, and the distance between the feet of the corresponding human object and the ground in the adjacent images after the target image is greater than 0.01 meters, then the acquisition time of the target image is the lifting start time.

[0057] Analyze each adjacent image to determine the position of the corresponding human object's legs in the image and the distance from the ground. Based on the positional relationship between the two, determine the distance between the legs and the ground. If the distance is greater than the preset leg-lifting distance, it is determined that the leg-lifting action is over. The acquisition time of the image that is first determined to be the end of the leg-lifting action is the leg-lifting end time.

[0058] For example, the preset leg-lifting distance is 0.3 meters, and the leg-lifting action can be judged to be completed when the distance is greater than 0.3 meters. The acquisition time of the first image judged to be the end of the leg-lifting action is the leg-lifting end time. If the distance between the legs of the corresponding human object and the ground in the adjacent images before the target image is less than 0.3 meters, and the distance between the legs of the corresponding human object and the ground in the target image is greater than 0.3 meters, the acquisition time of the target image is the lifting end time.

[0059] Step S303: Obtain the movement stroke of the toes according to the toe position corresponding to the start time of lifting and the toe position corresponding to the end time of lifting.

[0060] In this embodiment, the movement stroke of the toes may be the movement arc length from the toe position corresponding to the start time of the toes being lifted to the toe position corresponding to the end time of the toes being lifted.

[0061] Specifically, in the process of obtaining the movement stroke of the toes, the left and right sides of the human body object are taken as the x-axis, the direction from the head to the feet is taken as the y-axis, and the front and back of the human body object is taken as the z-axis to construct a three-dimensional space coordinate system of the human body object. In the initial state where the toes are not raised, the toe positions corresponding to the toes of both legs are at the origin of the three-dimensional coordinate system. In the state where the toes are raised, the toe positions corresponding to the raised toes are at a certain position in the space formed by the positive directions of the X-axis, the positive directions of the Y-axis, and the positive directions of the Z-axis. First, according to the limb feature data of the human body object, the leg length of the human body object and the distance between the raised toe position and the toe position of the coordinate origin are obtained. Then, with the length of both legs as the waist and the distance between the raised toe position and the toe position of the coordinate origin as the base, an isosceles triangle is constructed. According to the triangle interior angle sum theorem and the cosine theorem, the angle between the two legs in the state where the toes are raised is calculated. Finally, according to the calculated angle between the two legs and the leg length of the human body object, combined with the arc length calculation formula, the movement arc length of the toes is calculated, which is the movement stroke of the toes.

[0062] Step S304: Obtain the movement speed of the toes according to the movement stroke, the lifting start time and the lifting end time.

[0063] Step S305: If the movement speed is greater than or equal to the speed threshold, it is determined that the limb feature data of the human object meets the abnormal motion condition.

[0064] In this embodiment, the speed threshold may be a preset movement speed of the toes when raising the leg. For example, if the speed threshold is 0.6 meters per second, if the movement speed is greater than 0.6 meters per second, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0065] Specifically, in the process of obtaining the movement speed of the toes, first, the movement time of the toes being raised can be calculated based on the start time and the end time of the toes being raised, and then, the movement speed of the toes can be calculated based on the toes movement stroke and the movement time of the toes being raised, combined with the speed formula, and finally, the calculated movement speed is compared with a preset speed threshold. If the movement speed is greater than or equal to the speed threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0066] In this embodiment, for any human subject, when it is detected that the limb feature of the human subject meets the leg lifting feature, by detecting whether the movement speed of the toes is greater than or equal to the speed threshold, it is determined whether the limb feature data of the human subject meets the abnormal action condition. Through the above steps, it is realized to judge whether the action of the human subject is an abnormal action based on the size of the movement speed of the toes of the human subject, and then to judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of public safety hazards.

[0067] See also Figure 4 , is a flow chart of an abnormal event generation method provided in the third embodiment of the present invention. In the step S302 in the second embodiment, the movement stroke of the toes is obtained according to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time, which may include the following steps:

[0068] Step S401: determining the leg length from the limb feature data of the human subject.

[0069] Step S402: Calculate the first angle between the two legs based on the toe position corresponding to the start time of lifting and the toe position corresponding to the end time of lifting, combined with the leg length.

[0070] In this embodiment, the first angle between the two legs may be an angle formed by the intersection of two rays starting from the buttocks and with the direction of the legs pointing to the toes as the ray direction at the buttocks.

[0071] Specifically, in the process of calculating the first angle between the two legs of the human object on the image corresponding to the end time of lifting, first, the toe position corresponding to the start time of lifting and the toe position corresponding to the end time of lifting can be determined according to the limb feature data of the human object. Secondly, according to the toe position corresponding to the start time of lifting and the toe position corresponding to the end time of lifting, the straight-line distance between the toe position corresponding to the start time of lifting and the toe position corresponding to the end time of lifting is determined. Then, with the length of the two legs as the waist and the straight-line distance between the starting toe position and the ending toe position as the base, an isosceles triangle is constructed, wherein the vertex angle of the isosceles triangle is the first angle. Finally, according to the triangle interior angle sum theorem and the cosine theorem, the first angle between the two legs is calculated.

[0072] Step S403: obtaining the movement stroke of the toes according to the first angle and the leg length in combination with an arc length calculation formula.

[0073] In this embodiment, the movement stroke of the toes may be the length of the movement arc corresponding to the toes from the starting position to the ending position of the lifting.

[0074] Specifically, in the process of obtaining the movement stroke of the toes, first, a fan is constructed with the hip as the starting point, two rays pointing to the toes from both legs as the sides, the first angle formed by the intersection of the two rays at the hip as the vertex angle, and the movement stroke of the toes as the arc length. Then, according to the calculation formula of the fan arc length L: L = (πθ*R) / 180, where the radius R is the leg length and θ is the first angle, the arc length corresponding to the first angle is calculated, that is, the movement stroke of the toes is obtained. For example, if the first angle is 90° and the leg length is 0.8 meters, then according to the calculation formula of the fan arc length, the movement stroke of the toes can be calculated to be 0.4π.

[0075] In this embodiment, the first angle between the two legs is calculated, and the movement stroke of the toes is calculated according to the calculation formula of the first angle, leg length and arc length. Through the above steps, the movement stroke of the toes from the start time of the toes being lifted to the end time of the toes being lifted is calculated, which provides a data basis for the subsequent calculation of the movement speed of the toes.

[0076] See also Figure 5 , is a flow chart of an abnormal event generation method provided by Embodiment 4 of the present invention. In the above Embodiment 1, step S203 in detecting whether the limb feature data of each human object in at least one human object meets the abnormal action condition may further include the following steps:

[0077] S501: For any human subject, according to the limb feature data of the human subject, determine a second angle between a straight line where the thigh of the human subject's raised leg points toward the knee and a straight line where the trunk points toward the ground.

[0078] In this embodiment, the second angle may be the angle formed by the intersection of a ray starting from the buttocks and with the raised thigh pointing toward the knee as the ray direction and a ray with the torso pointing toward the ground as the ray direction at the buttocks.

[0079] S502: If the second angle is greater than the second angle threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0080] In this embodiment, the second angle threshold may be a preset second angle value. For example, if the second angle threshold is 60°, if the second angle is greater than 60°, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0081] For example, if the hip joint, knee and toe of the raised leg of the human subject are in the same straight line, in the process of detecting whether the second angle is greater than the second angle threshold, first, the straight-line distance between the toe position of the raised leg and the toe position of the leg that is not raised can be determined based on the toe position of the raised leg and the toe position of the leg that is not raised, and then, an isosceles triangle is constructed with the straight-line distance between the toe position of the raised leg and the toe position of the leg that is not raised as the base and the length of the leg as the waist, wherein the vertex angle of the isosceles triangle is the second angle, and finally, the second angle is calculated according to the triangle interior angle sum theorem and the cosine theorem, and the second angle is compared with the second angle threshold, and if the second angle is greater than the second angle threshold, it is determined that the limb feature data of the human subject meets the abnormal action condition.

[0082] In this embodiment, for any human subject, by detecting whether the second angle between the straight line where the human subject's raised leg points to the knee and the straight line where the torso points to the ground is greater than the second angle threshold, it is determined whether the limb feature data of the human subject meets the abnormal action condition. Through the above steps, it is realized to judge whether the action of the human subject is an abnormal action based on the size of the angle formed between the torso and the raised leg of the human subject, and then to judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of public safety hazards.

[0083] See also Figure 6 , is a flow chart of an abnormal event generation method provided in Embodiment 5 of the present invention. In the above Embodiment 1, step S203 of detecting whether the limb feature data of each human object in at least one human object meets the abnormal action condition may further include the following steps:

[0084] Step S601: for any human subject, according to the limb feature data of the human subject, determine the third angle between the legs of the human subject when the legs are not raised, and determine the fourth angle of the left leg bending and the fifth angle of the right leg bending in the human subject.

[0085] In this embodiment, the third angle can be the angle formed by two rays starting from the buttocks and with the legs pointing to the toes as the ray direction, intersecting at the buttocks when the legs of the human subject are not raised. The fourth angle and the fifth angle can be the angle formed by the rays starting from the knees and with the thighs pointing to the buttocks as the ray direction and the rays pointing from the calves to the ankles as the ray direction, intersecting at the knees.

[0086] Step S602: If the third angle is within the first preset angle range, and the fourth angle and the fifth angle are both within the second preset angle range, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0087] In this embodiment, the first preset angle range may be a preset third angle range, and the second preset angle range may be a preset fourth angle range and a preset fifth angle range.

[0088] Specifically, in the process of detecting whether the third angle is within the first preset angle range, first, when the legs of the human subject are not raised, the positions of the two ankles of the legs are determined according to the limb feature data of the human subject, and the straight-line distance between the two ankles is determined according to the positions of the two ankles of the legs. Then, an isosceles triangle is constructed with the length of the legs as the waist and the straight-line distance between the two ankles as the base, wherein the vertex angle of the isosceles triangle is the third angle. Finally, the third angle is calculated according to the triangle interior angle sum theorem and the cosine theorem, and the third angle is compared with the first preset angle range to detect whether the third angle is within the first preset angle range.

[0089] Specifically, in the process of detecting whether the fourth angle and the fifth angle are within the second preset angle range, first, when the legs of the human subject are bent, the positions of the two ankles of the legs and the position of the buttocks are determined according to the limb feature data of the human subject, and the straight-line distance from each ankle to the buttocks is determined according to the positions of the two ankles of the legs and the position of the buttocks. Then, for any leg, an isosceles triangle is constructed with the length of the thigh and the length of the calf as the waist and the straight-line distance from the ankle to the buttocks as the base, wherein the vertex angle of the isosceles triangle is the fourth angle or the fifth angle. Finally, according to the triangle interior angle sum theorem and the cosine theorem, the fourth angle and the fifth angle are calculated, and the fourth angle and the fifth angle are respectively compared with the second preset angle range to detect whether the fourth angle and the fifth angle are within the second preset angle range.

[0090] For example, if the first preset angle range is [30°, 45°], the second preset angle range is [45°, 170°], the calculated third angle is 30°, the fourth angle is 50°, and the fifth angle is 45°, then the third angle is within the first preset angle range, and the fourth angle and the fifth angle are both within the second preset angle range, and it is determined that the limb feature data of the human object meets the abnormal action condition.

[0091] In this embodiment, for any human subject, when the legs are not raised, by detecting whether the third angle between the legs is within the first preset angle range, and whether the fourth angle and the fifth angle of each bent leg are within the second preset angle range, it is determined whether the limb feature data of the human subject meets the abnormal action condition. Through the above steps, it is achieved to jointly determine whether the action of the human subject is an abnormal action by detecting the size of the angle formed between the legs and the size of the angle of the bent legs, and then to judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of public safety hazards.

[0092] See also Figure 7, is a flow chart of an abnormal event generation method provided in Embodiment 6 of the present invention. The abnormal event generation method provided in Embodiment 6 can be jointly judged with the abnormal event generation method provided in Embodiment 5 above. The abnormal event generation method can also include the following steps:

[0093] Step S701: For any human object, based on the limb feature data of the human object, determine the sixth angle between a straight line where the upper arm of any arm of the human object points toward the elbow and a straight line where the torso points toward the ground, determine the seventh angle of the arm bend, and determine the overlapping area between the arm and human objects other than the human object.

[0094] In this embodiment, the sixth angle can be the angle formed by the intersection of a ray with the shoulder as the starting point and a ray with the shoulder pointing to the elbow as the ray direction and a ray with the shoulder pointing to the ground as the ray direction at the shoulder. The seventh angle can be the angle formed by the intersection of a ray with the elbow as the starting point and a ray with the elbow pointing to the ankle as the ray direction and a ray with the elbow pointing to the shoulder as the ray direction at the elbow. The overlapping area is the overlapping area between the arm part of the human body object and the limb part of the human body object other than the human body object.

[0095] Step S702: If the sixth angle is within the third preset angle range, the seventh angle is within the fourth preset angle range, and the overlap area is less than the overlap area threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0096] In this embodiment, the third preset angle range can be a preset sixth angle range, the fourth preset angle range can be a preset seventh angle range, and the overlapping area threshold can be a preset area value of the overlap between the arm part of the human object and the limb part of the human object other than the human object.

[0097] Specifically, in the process of detecting whether the sixth angle is within the third preset angle range, first, the position of the elbow and the length of the upper arm are determined based on the limb feature data of the human subject. Secondly, with the shoulder as the starting point, a line segment equal to the length of the upper arm is taken on the torso where the shoulder points to the ground, and the straight-line distance from the elbow to the end point of the line segment is determined. Then, an isosceles triangle is constructed with the length of the upper arm and the length of the line segment as the waist and the straight line from the elbow to the end point of the line segment as the base, wherein the vertex angle of the isosceles triangle is the sixth angle. Finally, according to the triangle interior angle sum theorem and the cosine theorem, the sixth angle is calculated, and the sixth angle is compared with the third preset angle range to detect whether the sixth angle is within the third preset angle range.

[0098] Specifically, when the arm is bent, in the process of detecting whether the seventh angle is within the fourth preset angle range, first, the position of the ankle and the shoulder are determined according to the limb feature data of the human subject, and the straight-line distance from the ankle to the shoulder is determined according to the position of the ankle and the shoulder. Then, an isosceles triangle is constructed with the length of the upper arm and the length of the forearm as the waist and the straight-line distance from the ankle to the shoulder as the base, wherein the vertex angle of the isosceles triangle is the seventh angle. Finally, according to the triangle interior angle theorem and the cosine theorem, the seventh angle is calculated, and the seventh angle is compared with the fourth preset angle range to detect whether the seventh angle is within the fourth preset angle range.

[0099] Specifically, in the process of detecting whether the overlapping area of ​​the arms is smaller than the overlapping area threshold, the target detection algorithm can be used to locate the bounding box where the human object is located and the bounding box where the arm of the human object is located on the target image, and the overlapping area between the bounding box where the arm of the human object is located and the bounding box where the human objects other than the human object are located is calculated, that is, the overlapping area between the arm of the human object and the human objects other than the human object, and the overlapping area is compared with the overlapping area threshold to detect whether the overlapping area is smaller than the overlapping area threshold.

[0100] For example, if the third preset angle range is [45°, 85°], the fourth preset angle range is [30°, 160°], the overlap area threshold is 85, the calculated sixth angle is 50°, the seventh angle is 65°, and the overlap area is 40, then the sixth angle is within the third preset angle range, the seventh angle is within the fourth preset angle range, and the overlap area is less than the overlap area threshold, and it is determined that the limb feature data of the human object meets the abnormal action condition.

[0101] In this embodiment, for any human object, by detecting whether the sixth angle formed by the upper arm and the trunk is within the third preset angle range, whether the seventh angle of the arm bending is within the fourth preset angle range, and whether the overlap area is less than the overlap area threshold, it is determined whether the limb feature data of the human object meets the abnormal action condition. Through the above steps, it is achieved to jointly determine whether the action of the human object is an abnormal action by detecting the size of the angle formed by the upper arm and the trunk, the size of the angle of the arm bending, and the size of the overlap area between the arm and other human objects, and then to judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of public safety hazards.

[0102] See also Figure 8 , is a flow chart of an abnormal event generation method provided in Embodiment 7 of the present invention. The abnormal event generation method may further include the following steps:

[0103] Step S801: for any human object, determine the limb contact rate between the human object and any human object other than the human object according to the limb feature data of the human object, and determine the vertical distance between the head of the human object and the ground.

[0104] In this embodiment, the limb contact rate may be a ratio of an area of ​​direct contact between a limb part of a human subject and a limb part of another human subject to an area of ​​the limb of the human subject.

[0105] Specifically, in the process of determining the limb contact rate, first, the position of each human object is determined according to the limb feature data of the human object, and the straight-line distance between any two human objects is calculated. Secondly, each human object is segmented from the target image by an image segmentation algorithm, and the external contour of each human object is determined by an edge detection algorithm. Then, if the calculated straight-line distance between the two human objects is less than a preset distance threshold, the outer contour intersection area between the two human objects is calculated by comparing the distance of each pixel point in the outer contours of the two human objects. The contour intersection area is the direct contact area between the human object and other human objects. The limb contact rate can be calculated based on the direct contact area and the contour area of ​​the human object.

[0106] In the process of determining the vertical distance, the position of the head of the human subject and the position of the ground can be determined based on the limb feature data of the human subject, and the vertical distance from the head of the human subject to the ground can be determined based on the relationship between the two.

[0107] Step S802: If the limb contact rate is greater than the contact rate threshold, and the vertical distance is less than the second distance threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0108] In this embodiment, the contact rate threshold may be a preset contact rate between human objects, and the second distance threshold may be a preset vertical distance from the head of a human object to the ground. For example, if the contact rate threshold is 80%, and the second distance threshold is 1.20 m, if the calculated limb contact rate is 90%, and the calculated vertical distance is 0.90 m, then the calculated limb contact rate is greater than the contact rate threshold, and the calculated vertical distance is less than the second distance threshold, and it is determined that the limb feature data of the human object meets the abnormal action condition.

[0109] In this embodiment, by detecting whether the limb contact rate between the human object and other human objects other than the human object is greater than the contact rate threshold, and whether the vertical distance between the head of the human object and the ground is less than the second distance threshold, it is determined whether the limb feature data of the human object meets the abnormal action condition. Through the above steps, it is achieved to jointly judge whether the action of the human object is an abnormal action based on the contact rate between the human objects and the vertical distance between the head of the human object and the ground, and then judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of incidents endangering public safety.

[0110] See also Fig. 9 , is a flow chart of an abnormal event generation method provided by Embodiment 8 of the present invention. In the above Embodiment 1, step S203 in detecting whether the limb feature data of each human object in at least one human object meets the abnormal action condition may further include the following steps:

[0111] Step S901: Determine the facial orientation of each human object according to the limb feature data of each human object.

[0112] Step S902: for any human object, determine a target human object whose face orientation is opposite to that of the human object from at least one human object, and regard the human object and the target human object as a human pair.

[0113] In this embodiment, the facial orientation may be the facial orientation direction of the human body object, for example, the facial orientation may be the left, left front, front, right front, and right of the human body object. A human body pair may refer to two human body objects, namely a first human body and a second human body, wherein the facial orientation of the first human body points to the second human body, and the facial orientation of the second human body points to the first human body.

[0114] Specifically, in the process of determining the facial orientation of each human object and determining the human pair, the facial orientation of each human object can be obtained by learning vector quantization neural network recognition. First, feature information such as the eye position and orientation angle of each human object is extracted according to the edge operator, and then the extracted feature information is input into the learning vector quantization neural network, and the facial orientation of each human object in the target image is obtained through classification and recognition. Finally, the facial orientation of each human object is compared to determine that two human objects with the same facial orientation are a human pair, and the same label is given to the two human objects in the same human pair to facilitate subsequent analysis.

[0115] Step S903: Calculate the interval distance between two human objects in the human body pair.

[0116] Step S904: If the interval distance is less than the first distance threshold, the facial orientations of other human objects other than the human body pair are obtained.

[0117] In this embodiment, the interval distance can be the straight-line distance between the two human objects in the human body pair, and the first distance threshold can be a preset interval distance. For example, if the first distance threshold is 0.05 meters, if the calculated interval distance between the two human objects in the human body pair is 0.02 meters, then the interval distance is less than the first distance threshold.

[0118] Step S905: Calculate the proportion of other human objects whose facial orientations are human pairs among all other human objects.

[0119] Specifically, in the process of calculating the proportion of other human objects with facial orientations of human pairs among all other human objects, first, the facial orientations of other human objects other than all human pairs are obtained, then, the number of other human objects other than all human pairs is determined, and the number of other human objects with facial orientations of human pairs is determined, and finally, based on the number of other human objects and the number of other human objects with facial orientations of human pairs, the proportion of other human objects with facial orientations of human pairs among all other human objects is calculated.

[0120] Step S906: If the proportion is greater than the preset proportion threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

[0121] In this embodiment, the preset proportion threshold can be the proportion of other human objects with a preset facial orientation of the human body among all other human objects. For example, if the preset proportion threshold is 85% and the calculated proportion is 90%, the calculated proportion is greater than the preset proportion threshold, and it is determined that the limb feature data of the human object meets the abnormal action condition.

[0122] In this embodiment, by determining a human body pair and detecting whether the interval distance between two human body objects in the human body pair is less than a first distance threshold, if the interval distance is less than the first distance threshold, detecting whether the proportion of other human body objects whose faces are facing the human body pair among all other human body objects is greater than a preset proportion threshold, it is determined whether the limb feature data of the human body object meets the abnormal action condition. Through the above steps, it is realized to jointly determine whether the action of the human body object is an abnormal action based on the facial orientation of the human body object, the distance between the human body objects, and the proportion, and then to judge the abnormal situation, thereby improving the accuracy of abnormal situation detection and effectively preventing the occurrence of public safety hazards.

[0123] See also Fig.10, is an abnormal event generation device provided by Embodiment 9 of the present invention, and the abnormal event generation device corresponds to the abnormal event generation method in the above embodiment. The abnormal event generation device includes a first acquisition module 101, an analysis module 102, a first detection module 103 and a generation module 104. Each functional module is described in detail as follows:

[0124] A first acquisition module 101 is used to acquire a target image, where the target image is an image that meets a crowd gathering condition and includes at least one human object;

[0125] An analysis module 102 is used to perform limb analysis on each human object in the target image to obtain limb feature data of each human object;

[0126] A first detection module 103, used to detect whether the limb feature data of each human object in the at least one human object meets the abnormal action condition;

[0127] The generating module 104 is used to generate an abnormal event when it is detected that the limb feature data of all human objects meets the abnormal action condition.

[0128] Optionally, the first detection module 103 includes:

[0129] A second acquisition unit is used for acquiring adjacent images before and after the target image for any human object when detecting that the limb feature data of the human object meets the leg lifting feature;

[0130] A third acquisition unit is used to obtain the lifting start time and lifting end time corresponding to the leg lifted by the human subject according to the acquisition time of all adjacent images and the acquisition time of the target image;

[0131] A fourth acquisition unit, configured to obtain a movement stroke of the toe according to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time;

[0132] The first calculation unit is used to obtain the movement speed of the toes according to the movement stroke, the lifting start time and the lifting end time.

[0133] The first determination unit is used to determine that the limb feature data of the human object meets the abnormal action condition if the movement speed is greater than or equal to a speed threshold.

[0134] Optionally, the fourth obtaining unit includes:

[0135] A second determination subunit is used to determine the leg length from the limb feature data of the human subject;

[0136] A second calculation subunit is used to calculate a first angle between the two legs according to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time, combined with the leg length;

[0137] The third calculation subunit is used to obtain the movement stroke of the toe according to the first angle and the leg length in combination with an arc length calculation formula.

[0138] Optionally, the first detection module 103 further includes:

[0139] A third determination unit is used to determine, for any human subject, a second angle between a straight line where the thigh of the human subject's raised leg points in the direction of the knee and a straight line where the trunk points in the direction of the ground according to the limb feature data of the human subject;

[0140] The fourth determining unit is used to determine that the limb feature data of the human object meets the abnormal action condition if the second angle is greater than a second angle threshold.

[0141] Optionally, the first detection module 103 further includes:

[0142] a fifth determining unit, for determining, for any human subject, a third angle between the legs of the human subject when the legs are not lifted, and a fourth angle when the left leg is bent and a fifth angle when the right leg is bent in the human subject, according to the limb feature data of the human subject;

[0143] a sixth determining unit, configured to determine, based on the limb feature data of the human subject, a sixth angle between a straight line where the upper arm of any arm of the human subject points to the direction of the elbow and a straight line where the trunk points to the direction of the ground, determine a seventh angle of the bending of the arm, and determine an overlapping area between the arm and a human subject other than the human subject;

[0144] The seventh determination unit is used to determine that the limb feature data of the human object meets the abnormal action condition if the third angle is within the first preset angle range, and the fourth angle and the fifth angle are both within the second preset angle range, and / or the sixth angle is within the third preset angle range, the seventh angle is within the fourth preset angle range, and the overlapping area is less than the overlapping area threshold.

[0145] Optionally, the first detection module 103 further includes:

[0146] an eighth determination unit, configured to determine, for any human subject, a limb contact rate between the human subject and any human subject other than the human subject, and a vertical distance between the head of the human subject and the ground according to the limb feature data of the human subject;

[0147] The ninth determination unit is used to determine that the limb feature data of the human object meets the abnormal action condition if the limb contact rate is greater than the contact rate threshold and the vertical distance is less than the second distance threshold.

[0148] Optionally, the first detection module 103 further includes:

[0149] The ninth determining unit is used to determine the facial orientation of each human object according to the limb feature data of each human object.

[0150] a tenth determining unit, configured to determine, for any human object, a target human object having a face orientation opposite to that of the human object from among the at least one human object, and to regard the human object and the target human object as a human pair;

[0151] A fourth calculation unit, used for calculating the interval distance between two human objects in the human body pair;

[0152] a second detection unit, configured to obtain the facial orientations of other human objects other than the human body pair if the interval distance is less than the first distance threshold;

[0153] A fifth calculation unit, configured to calculate a proportion of other human objects whose facial orientation is the human pair among all other human objects;

[0154] An eleventh determining unit is used to determine that the limb feature data of the human object meets an abnormal action condition if the proportion is greater than a preset proportion threshold.

[0155] For the specific definition of the event analysis device, please refer to the definition of the abnormal event generation method above, which will not be repeated here. Each module in the above-mentioned abnormal event generation device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0156] Fig.11 This is a schematic diagram of the structure of a terminal device provided in Embodiment 10 of the present invention. Fig.11 As shown, the terminal device of this embodiment includes: at least one processor ( Fig.11 Only one is shown in the figure), a memory, and a computer program stored in the memory and executable on at least one processor, wherein when the processor executes the computer program, the steps in any of the above-mentioned abnormal event generation method embodiments are implemented.

[0157] The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that Fig.11 It is only an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include a network interface, a display screen, and an input device.

[0158] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0159] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory may be the memory of the terminal device, and the internal memory provides an environment for the operation of the operating system and the computer-readable instructions in the readable storage medium. The readable storage medium may be the hard disk of the terminal device, and in other embodiments, it may also be an external storage device of the terminal device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal device. Further, the memory may also include both an internal storage unit of the terminal device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of a computer program, etc. The memory may also be used to temporarily store data that has been output or is to be output.

[0160] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device capable of carrying computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0161] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by a computer program product. When the computer program product runs on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiment when executing it.

[0162] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0163] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0164] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0165] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0166] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents, and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for generating an abnormal event, characterized in that: The abnormal event generation method comprises the following steps: Acquire a target image, where the target image is an image that meets a crowd gathering condition and includes at least one human object; Performing limb analysis on each human object in the target image to obtain limb feature data of each human object; Detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition; When it is detected that among the limb feature data of all human objects, limb feature data satisfies the abnormal action condition, an abnormal event is generated.

2. The abnormal event generation method according to claim 1, characterized in that: The detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition includes: For any human object, when it is detected that the limb feature data of the human object meets the leg lifting feature, an adjacent image of the target image is acquired; Obtaining the lifting start time and lifting end time corresponding to the leg lifted by the human subject according to the acquisition time of all adjacent images and the acquisition time of the target image; According to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time, a movement stroke of the toe is obtained; According to the movement stroke, the lifting start time and the lifting end time, the movement speed of the toes is obtained; If the movement speed is greater than or equal to the speed threshold, it is determined that the limb feature data of the human object meets the abnormal movement condition.

3. The abnormal event generation method according to claim 2, characterized in that: The step of obtaining the movement stroke of the toes according to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time comprises: Determining leg length from limb feature data of the human subject; Calculate the first angle between the two legs according to the toe position corresponding to the lifting start time and the toe position corresponding to the lifting end time in combination with the leg length; The movement stroke of the toe is obtained according to the first angle and the leg length in combination with an arc length calculation formula.

4. The abnormal event generation method according to claim 1, characterized in that: The detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition also includes: For any human subject, according to the limb feature data of the human subject, determine a second angle between a straight line where the thigh of the human subject's raised leg points toward the knee and a straight line where the trunk points toward the ground; If the second angle is greater than a second angle threshold, it is determined that the limb feature data of the human object meets an abnormal action condition.

5. The abnormal event generation method according to claim 1, characterized in that: The detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition also includes: For any human subject, according to the limb feature data of the human subject, determine a third angle between the legs of the human subject when the legs are not lifted, and determine a fourth angle when the left leg of the human subject is bent and a fifth angle when the right leg is bent; According to the limb feature data of the human subject, determine a sixth angle between a straight line where the upper arm of any arm of the human subject points to the elbow and a straight line where the trunk points to the ground, determine a seventh angle where the arm is bent, and determine an overlapping area between the arm and a human subject other than the human subject; If the third angle is within the first preset angle range, and the fourth angle and the fifth angle are both within the second preset angle range, and / or the sixth angle is within the third preset angle range, the seventh angle is within the fourth preset angle range, and the overlapping area is less than the overlapping area threshold, then it is determined that the limb feature data of the human object meets the abnormal action condition.

6. The abnormal event generation method according to claim 1 or 5, characterized in that: The detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition also includes: For any human subject, determining a limb contact rate between the human subject and any human subject other than the human subject, and determining a vertical distance between the head of the human subject and the ground according to the limb feature data of the human subject; If the limb contact rate is greater than the contact rate threshold, and the vertical distance is less than the second distance threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

7. The abnormal event generation method according to claim 1, characterized in that: The detecting whether the limb feature data of each human object in the at least one human object meets the abnormal action condition also includes: Determining the facial orientation of each human subject according to the limb feature data of each human subject; For any human object, determining a target human object having a face orientation opposite to that of the human object from the at least one human object, and treating the human object and the target human object as a human pair; Calculating the separation distance between two human objects in the human body pair; If the interval distance is less than a first distance threshold, acquiring the facial orientations of other human objects other than the human body pair; Calculate the proportion of other human objects whose facial orientation is the human pair among all other human objects; If the proportion is greater than a preset proportion threshold, it is determined that the limb feature data of the human object meets the abnormal action condition.

8. An abnormal event generating device, characterized in that: The abnormal event generating device comprises: A first acquisition module is used to acquire a target image, wherein the target image is an image that meets a crowd gathering condition and includes at least one human object; An analysis module, used for performing limb analysis on each human object in the target image to obtain limb feature data of each human object; A first detection module, used to detect whether the limb feature data of each human object in the at least one human object meets the abnormal action condition; The generating module is used to generate an abnormal event when it is detected that the limb feature data of all human objects meets the abnormal action condition.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the abnormal event generating method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the abnormal event generation method according to any one of claims 1 to 7 are implemented.

Images