Human body tumble detection method, computer readable storage medium and electronic device

A detection method and technology of human body, applied in computer parts, calculation, character and pattern recognition, etc., can solve the problem of high false positive rate of fall detection technology

Pending Publication Date: 2021-07-16
深圳市真元天芮智能软件有限公司
0 Cites 3 Cited by

AI-Extracted Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a human body fall detection method, a computer-readable storage...
View more

Abstract

The invention relates to a machine vision technology, discloses a human body tumble detection method, a computer readable storage medium and an electronic device, and mainly solves the problem that an existing tumble detection technology is relatively high in misjudgment rate. According to the invention, the skeleton key points of the human body target in the monitoring video are detected in real time, and the skeleton key points of the human body target in the continuous preset number of frames before the current moment are analyzed to judge whether the human body target falls down, so that the detection of the human body falling process is realized from a dynamic angle, and then whether the human body falls down can be accurately judged.

Application Domain

Biometric pattern recognition

Technology Topic

EngineeringHuman body +2

Image

  • Human body tumble detection method, computer readable storage medium and electronic device

Examples

  • Experimental program(1)

Example Embodiment

[0060] In order to make the objects, technical solutions, and advantages of the present invention, the present invention will be further described in detail below in connection with the embodiments and drawings.
[0061] One human fall detection method provided by the present invention is mainly to detect and analyze the characteristics and changes in the bone key point of the human body, thereby judge whether or not the human body target has fallen. Overall, the human fall detection method includes the following steps:
[0062] Step S1: Real-time detection monitoring of human body objectives;
[0063] Step S2: Real-time detection of bone key points of the human objective;
[0064] Step S3: Decades whether the human body target has fallen according to the bone key of the human body target in the continuous preset quantity frame before the current time.
[0065] In step S1, the human body target detection and tracking algorithm can be used to detect and track it here.
[0066] In step S2, 18 bone key points of the human body target can be selected for detection. These 18 bone key points are: 0- Nose, 1-neck, 2-right shoulders, 3-right elbow, 4-right wrist, 5- Left shoulder, 6-left elbow, 7-left wrist, 8-right hip, 9-right knee, 10-right ankle, 11-left hip, 12-left knee, 13-left ankle, 14-right eye, 15-left eye 16-Right ear, 17-left ear. OpenPOSE can be used to detect bone key points. OpenPOSE is a open source, facial expression, finger motion, etc., which can achieve human actions, facial expressions, finger motion, etc., which can achieve human motion, facial expressions, finger movement, etc., based on convolutional neural network and open source libraries developed by convolutional neural network and supervision and learning and Caffe. Track, suitable for single and multiplayers, with excellent robustness.
[0067] Step S3 is the main innovation of the present invention. Step S3 can be implemented in a variety of ways, in a different manner, step S3 may include different specific steps. The various implementations of step S3 are as follows.
[0068] In the first mode, step S3 includes:
[0069] Determine the center of gravity of human body in each frame according to the bone key point of the human target in each frame;
[0070] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0071] When the falling distance is greater than the first threshold, and the average drop speed is greater than the second threshold, it is determined that the human target falls.
[0072] The first way is mainly determined to achieve the decline in the human body according to the descending distance and falling speed of the center of gravity of the human body. In general, when the human body falls, its center of gravity will be reduced to the height of the ground, and its center of gravity is lowered from the original height to the height of the ground, and the distance from the center of gravity is basically the hip joint of the hip joint of the body. . The height of the hip joint distance from the hip joint distance from the hip joint of the hip joint is calculated according to the bone key of the human body, and the height is calculated as the first threshold. Alternatively, according to the bone key of the human body, the middle point of the two hip joints of the body target is calculated from the midpoint of the middle point of the two ankle joints, and the height is used as the first threshold. At the same time, in the falling speed, since the human body falls is not autonomous, the flying speed of its center of gravity is higher than usually squatting or sitting down during the human fall, and you can set it down or sit down next to us. The center of gravity has a higher value (for example, 1 m / sec or 2 m / sec) as the second threshold. The first way only analyzes the change of the center of gravity of the human body to determine whether the human body target has fallen, and the detection of the human falling process is also taken into account, and the basic accurate test results can be obtained, while only the human object is required The focus of the center of gravity, the amount of calculation is small, can save computational resources, and improve the detection speed.
[0073] In the second method, step S3 includes:
[0074] Determine the gravity and support surface of the human body objects in each frame according to the bone key point of the human body target in each frame;
[0075] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0076] Depending on the center of gravity of the human body objective in each frame, it is judge whether the center of gravity of the human body target in each frame falls within its support surface;
[0077] When the falling distance is greater than the first threshold, and the average drop rate is greater than the second threshold, and when the center of gravity of the human body target in each frame is not in its support surface, the human body target has fallen.
[0078] The second way is different from the first way in that the second method is added to the analysis factors of the center of gravity line and the support surface on the basis of the first method. The center of gravity of the human body refers to a straight line from the center of gravity to the center of gravity. It is an important basis and auxiliary line for analyzing the human movement. The support surface of the human body refers to the range of all the outermost contact points in each of the respective contact points of the human body, such as the ground), so the support surface is not an unlimited plane. In general, when the human body occurs or is about to fall, its center of gravity will fall outside its support surface, and the second way further according to the center of gravity in each frame, it is judged whether or not the center of gravity of the human body in each frame is falling In the support surface, this can be judged to make a judgment on whether the human body target in each frame is determined. In the second method, in the premise of satisfying the descent distance of the center of gravity in the first mode and the detection result of the average decrease speed, it is also necessary to satisfy the center of gravity in each frame (ie, the center of the human body in each frame). None of its support surface) This test result is determined that the human body target has fallen, and it is possible to further improve the detection accuracy of the human fall than the first method.
[0079] In the third method, step S3 includes:
[0080] Determine the longitudinal centerline of the human body target in each frame according to the bone key point of the human body target in each frame;
[0081] Calculate the angle between the longitudinal centerline and the ground in each frame;
[0082] According to the angle between the longitudinal centerline and the ground, the angle between the longitudinal centerline and the ground is calculated, and the reduction amount and reduction rate of the angle between the predetermined quantity frames are calculated;
[0083] When the amount is larger than the third threshold, and the reduction rate is larger than the fourth threshold, and when the angle between the longitudinal centerline and the ground planes in each frame is less than the fifth threshold, it is determined that the human body target has fallen.
[0084] The third way is different from the first mode and second mode, and the third manner determines whether the human body has fallen by calculating the angle between the vertical centerline and the ground plane in the frame, and analyzes the change in the angle. The longitudinal centerline of the human body refers to the line of the two parts of the body contour into the left and right symmetrical. For example, a triangular shape can be formed in a human body 1-neck, 8-right hip, 11-left hip three points, calculating a middle line of a 1-neck, and the center line is used as the longitudinal centerline of the human body. Selecting these three points to form a triangle and calculate the advantages of the human body vertical center line, regardless of the posture of the human body, the positional relationship between these three points is relatively fixed, so through these three Talk to determine the vertical centerline of the human body is relatively accurate. In general, when the human body falls, the angle between the longitudinal centerline and the ground will decrease. Ideally, when the human body is standing normally, the angle between the longitudinal centerline and the ground should be 90 degrees, and when the human body is flat, the angle between the longitudinal centerline and the surface of the ground should be 0 degrees. When the human body falls, the angle between the longitudinal center of gravity and the ground should be less than 90 degrees, and it is close to 0 degrees. It can set a suitable degree, which meets the angle between the human body vertical center line and the ground plane. This degree is used as the fifth threshold. At the same time, in general, the velocity of the human body will be faster than the normal movement of the body, and the body falls, and the reduction rate of the angle between the vertical center line and the ground is also sitting than the human body. The reduction rate of the angle between the longitudinal centerline and the ground is large when the normal action is lie down, and the degree of reduction rate is higher than the angle of the normal action to sit down or lying down or lying down. For example, 45 degrees / second or 90 degrees / sec) as the fourth threshold. At the same time, when the human body falls, the reduction in the longitudinal centerline and the surface between the preset quantity frames should also be larger than usual, and the angle is performed. In this market, a suitable value (e.g., 60 degrees) can be provided as the third threshold. The third way is different from the first way in that the third means reflects the human falling process by the change in the angle between the human vertical centerline and the ground, and the first way is through the changes of human center. Reflect the fall of the human body.
[0085] In the fourth way, step S3 includes:
[0086] Determine the center of gravity and minimum external rectangle in each frame according to the bone key point of the human body target in each frame;
[0087] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0088] Calculate the width ratio of minimum external rectangles in each frame;
[0089] When the falling distance is greater than the first threshold, and the average drop rate is greater than the second threshold, and when the aspect ratio is greater than the sixth threshold, it is determined that the human body target falls.
[0090]The difference between the fourth way and the first method is that the analysis factor of the minimum external rectangle is added based on the first way. When the human body is in the fall state and the non-dropped state, the width ratio of the minimum external rectangle of the human body target captured by the same fixed camera is different, and the width ratio of the minimum external rectangle means the left and right width of the minimum external rectangle. High refers to the upper and lower height of the minimum external rectangle. When the body is in a normal state, the width ratio of its minimum external rectangle is smaller, and when the human body falls, the width ratio of its minimum external rectangle is large, and the minimum external rectangular rectangle of the human body target is detected in real time. The aspect ratio can reflect whether the human body is falling from the other side, and a suitable threshold (e.g., 1/2) can be set, and when it is smaller than this value, it is determined that the human body is non-falling state, and when it is greater than this value, it is determined that the human body falls. status. The minimum external rectangle determined according to the 18 bone key points of the human body target can be directly determined as the minimum external rectangle of the human body. In the fourth manner, in the premise of satisfying the descending distance of the center of gravity in the first mode and the detection result of the average decrease speed, it is also necessary to meet the width ratio of the minimum external rectangle of the human body in each frame is greater than the sixth threshold. A test result is determined that the human body target has fallen, and it is possible to further improve the detection accuracy of the human body than the first method.
[0091] In the fifth mode, step S3 includes:
[0092] Determine the center of gravity, support surface and minimum external rectangle in each frame according to the bone key of the human body target in each frame;
[0093] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0094] Depending on the center of gravity of the human body objective in each frame, it is judge whether the center of gravity of the human body target in each frame falls within its support surface;
[0095] Calculate the width ratio of minimum external rectangles in each frame;
[0096] When the falling distance is greater than the first threshold, and the average drop rate is greater than the second threshold, and the center of gravity of the human body target in each frame does not pass through its support surface, and when the aspect ratio is larger than the sixth threshold, it is determined that the human body target falls. .
[0097] The fifth way is to add a minimum external rectangular analysis factor on the second method, and can further improve the detection accuracy of the human fall than the second method. The analysis of the minimum external rectangle of the human body can refer to the description in the fourth embodiment described above, and will be described herein.
[0098] In the sixth method, step S3 includes:
[0099] Determine the longitudinal centerline and minimum external rectangle of the human body target in each frame according to the bone key of the human body target in each frame;
[0100] Calculate the angle between the longitudinal centerline and the ground in each frame;
[0101] According to the angle between the longitudinal centerline and the ground, the angle between the longitudinal centerline and the ground is calculated, and the reduction amount and reduction rate of the angle between the predetermined quantity frames are calculated;
[0102] Calculate the width ratio of minimum external rectangles in each frame;
[0103] When the amount is larger than the third threshold, and the reduction rate is larger than the fourth threshold, and the angle between the longitudinal centerline and the ground between the respective frames is less than the fifth threshold, and when the width is larger than the sixth threshold, the human body target is determined. Appearance.
[0104] The sixth way is to add a minimum external rectangle based on the third way, and can further improve the detection accuracy of the human fall than the third method. The analysis of the minimum external rectangle of the human body can refer to the description in the fourth embodiment described above, and will be described herein.
[0105] In seventh mode, step S3 includes:
[0106] Determine the center of gravity, support surface and longitudinal centerline of human body objects in each frame according to the bone key of the human body target in each frame;
[0107] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0108] Depending on the center of gravity of the human body objective in each frame, it is judge whether the center of gravity of the human body target in each frame falls within its support surface;
[0109] Calculate the angle between the longitudinal centerline and the ground in each frame;
[0110] According to the angle between the longitudinal centerline and the ground, the angle between the longitudinal centerline and the ground is calculated, and the reduction amount and reduction rate of the angle between the predetermined quantity frames are calculated;
[0111] When the falling distance is greater than the first threshold, and the average drop rate is greater than the second threshold, and the center of gravity of the human body target in each frame does not pass through the support surface, and the amount is larger than the third threshold, and the reduction rate is greater than the first The four threshold, and when the angle between the longitudinal centerline and the ground is less than the fifth threshold, it is determined that the human body target falls.
[0112] The seventh way is equivalent to the combination of the second method and the third method, so that the center of gravity, center of gravity, support surface, and longitudinal centerline are integrated to determine whether the human body has fallen, only when various factors indicate When the human target falls, it is determined that the human body target has fallen.
[0113] In the eighth mode, step S3 includes:
[0114] Determine the center of gravity, support surface, longitudinal centerline and minimum external rectangle of human body target in each frame according to the bone key of the human body target in each frame;
[0115] According to the center of gravity of the human body objective in each frame, the descent distance and the average decrease rate of the center of gravity corresponding to the preset quantity frame is calculated;
[0116] Depending on the center of gravity of the human body objective in each frame, it is judge whether the center of gravity of the human body target in each frame falls within its support surface;
[0117] Calculate the angle between the longitudinal centerline and the ground in each frame;
[0118] According to the angle between the longitudinal centerline and the ground, the angle between the longitudinal centerline and the ground is calculated, and the reduction amount and reduction rate of the angle between the predetermined quantity frames are calculated;
[0119] Calculate the width ratio of minimum external rectangles in each frame;
[0120] When the falling distance is greater than the first threshold, and the average drop rate is greater than the second threshold, and the center of gravity of the human body target in each frame does not pass through the support surface, and the amount is larger than the third threshold, and the reduction rate is greater than the first The four threshold, and the angle between the longitudinal centerline and the ground between each frame is less than the fifth threshold, and when the aspect is greater than the sixth threshold, it is determined that the human body target falls.
[0121] The eighth way is to add a minimum external rectangular analysis factor on the basis of seventh way, and can further improve the detection accuracy of the human body in the seventh mode. The analysis of the minimum external rectangle of the human body can refer to the description in the fourth embodiment described above, and will be described herein.
[0122] The present invention also provides a computer readable storage medium that stores a computer program capable of being executed by the processor, and the computer program is executed by the processor, and any of the above human fall detection methods is implemented.
[0123] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and can run on the processor, and when the processor performs a computer program, any of the above human fall detection methods.
[0124] The above embodiments are merely preferred embodiments, and are not intended to limit the scope of the invention, any modifications, equivalents, and improvements made within the spirit and principles of the present invention, should be included in the scope of protection of the present invention. .

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products