Fall detection method, fall protection vest and electronic device

Abstract

The embodiments of the present application provide a fall detection method, a fall protection vest and an electronic device, which belong to the field of image processing technology. The method comprises: acquiring an environmental image and acceleration sensor data; wherein the environmental image and the acceleration sensor data correspond to the motion state of the target object at at least two detection moments; extracting line features from the environmental image at each detection moment to obtain corresponding pattern line features; calculating the rotation angular velocity of the target object based on the pattern line features corresponding to at least two detection moments; judging whether the target object falls based on the rotation angular velocity and the acceleration sensor data. The embodiments of the present application can improve the accuracy of fall detection.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a fall detection method, a fall protection vest and an electronic device. Background Art

[0002] Falls in the elderly can cause serious physical injuries, including fractures and head injuries. Due to reduced bone density and decreased physical function, the elderly are more susceptible to injury, and the recovery process may be slower and more difficult.

[0003] In the related art, there is a technology that uses a wearable acceleration sensor to detect when a user is about to fall, and then opens the airbag to prevent the user from getting injured. However, this technology has the problem of low accuracy: when the user is performing daily life activities with large acceleration, it is mistakenly judged that the user is about to fall, but the user actually does not show a tendency to fall, and the system mistakenly opens the airbag, which will result in the waste of a gas cylinder worth hundreds of yuan. Moreover, after the airbag is mistakenly opened, it may scare the user and cause additional injuries, and the expanded airbag will cause inconvenience to the user's subsequent actions. Therefore, how to improve the accuracy of fall detection has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a fall detection method, a fall protection vest and an electronic device, aiming to improve the accuracy of fall detection.

[0005] To achieve the above object, a first aspect of an embodiment of the present application provides a fall detection method, the method comprising:

[0006] Acquire an environmental image and acceleration sensing data; wherein the environmental image and the acceleration sensing data correspond to the motion state of the target object at at least two detection moments;

[0007] Extracting line features from the environment image at each detection moment to obtain corresponding pattern line features;

[0008] Calculating the rotational angular velocity of the target object according to the pattern line features corresponding to at least two of the detection moments;

[0009] Based on the rotational angular velocity and the acceleration sensor data, determine whether the target object falls

[0010] In some embodiments, extracting line features from the environment image at each detection moment to obtain corresponding pattern line features includes:

[0011] Performing denoising on the environmental image to obtain a pure image;

[0012] Performing edge detection on the clean image and performing non-maximum suppression processing to obtain an edge binary image;

[0013] Performing edge cutting on the edge binary image based on a predetermined cutting parameter to obtain a preprocessed image;

[0014] Line features are extracted from the preprocessed image to obtain the pattern line features.

[0015] In some embodiments, the pattern line features include ellipse feature data; the line feature extraction for the environment image at each detection moment to obtain the corresponding pattern line features includes:

[0016] Determining a first number of edge points included in the environment image;

[0017] For the first number of edge points, establish edge point pairs according to any two of the edge points to obtain a second number of edge point pairs;

[0018] Generate a target ellipse pattern for each edge point pair, and determine an ellipse voting value corresponding to the target ellipse pattern;

[0019] Based on the target ellipse pattern and the ellipse voting value corresponding to the target ellipse pattern, the corresponding ellipse feature data is obtained.

[0020] In some embodiments, generating a target ellipse pattern for each edge point pair and determining an ellipse voting value corresponding to the target ellipse pattern includes:

[0021] Determining the major axis endpoints of the elliptical pattern according to the edge point pairs;

[0022] Determine the edge points other than the edge point pairs among the first number of edge points as third edge points;

[0023] Determine a plurality of candidate ellipse patterns according to the edge point pair and each of the third edge points, and calculate a candidate minor axis length of each of the candidate ellipse patterns;

[0024] Determining the ellipse voting value of the corresponding candidate ellipse pattern according to the number of occurrences of each candidate short axis length;

[0025] The candidate ellipse pattern whose ellipse voting value meets the first preset condition is determined as the target ellipse pattern.

[0026] In some embodiments, the pattern line features further include straight line feature data, and the line feature extraction is performed on the environment image at each detection moment to obtain the corresponding pattern line features, including:

[0027] Performing Hough transform processing on the first number of edge points to obtain a plurality of candidate straight line patterns and straight line voting values ​​corresponding to the candidate straight line patterns;

[0028] When the straight line voting value meets the second preset condition, the corresponding candidate straight line pattern is determined as the target straight line pattern;

[0029] Based on the target straight line pattern and the corresponding straight line voting value, the corresponding straight line feature data is obtained.

[0030] In some embodiments, the pattern line features further include circular feature data, and the line feature extraction is performed on the environment image at each detection moment to obtain the corresponding pattern line features, including:

[0031] Selecting two of the environment images at the detection moments from the environment images at each detection moment; wherein the two detection moments are respectively a first detection moment and a second detection moment;

[0032] Performing Hough transform processing on the edge points at the first detection moment to obtain a plurality of first candidate circular patterns and corresponding first circular voting values;

[0033] Performing Hough transform processing on the edge points at the second detection moment to obtain a plurality of second candidate circular patterns and corresponding second circular voting values;

[0034] When the first circular voting value and the second circular voting value meet the third preset condition, and the relative circle center distance meets the fourth preset condition, the first target circular pattern at the first detection moment and the second target circular pattern at the second detection moment are obtained; wherein the relative circle center distance represents the distance between the center of the first candidate circular pattern corresponding to the first circular voting value and the center of the second candidate circular pattern corresponding to the second circular voting value;

[0035] Based on the first target circular pattern and the second target circular pattern, obtaining a circle center change angle;

[0036] Obtaining a circular pattern voting mean value according to a first circular voting value corresponding to the first target circular pattern and a second circular voting value corresponding to the second target circular pattern;

[0037] The circular feature data is obtained based on the circle center variation angle, the circular pattern voting mean, the first target circular pattern and the second target circular pattern.

[0038] In some embodiments, the step of calculating the rotational angular velocity of the target object based on the pattern line features corresponding to at least two of the detection moments, taking two of the detection moments as the first target moment and the second target moment, specifically includes:

[0039] Obtaining first deflection data according to the ellipse feature data and the straight line feature data at the first target moment;

[0040] Obtaining second deflection data according to the ellipse feature data and the straight line feature data at the second target moment;

[0041] The rotational angular velocity of the target object is calculated based on the first deflection data, the second deflection data, and the circular feature data.

[0042] In some embodiments, extracting line features from the environment image at each detection moment to obtain corresponding pattern line features further includes:

[0043] Recording the number of the target elliptical patterns at the detection moment as a first determination number;

[0044] recording the number of the target straight line patterns at the detection moment as a second determination number;

[0045] recording the number of the first target circular pattern or the second target circular pattern at the detection moment as a third determination number;

[0046] Calculating image feature values ​​based on the first determination number, the second determination number, and the third determination number to obtain the image feature value at the detection moment;

[0047] Wherein, after judging whether the target object has fallen based on the rotation angular velocity and the acceleration sensor data, the method further comprises:

[0048] Based on the image feature value at each of the detection moments, a quantitative feature change value on a preset time series is obtained; wherein the preset time series is composed of a third number of the detection moments;

[0049] When the quantity characteristic change value satisfies a fifth preset condition at a third target moment in the preset time sequence, and the quantity characteristic change value satisfies a sixth preset condition after the third target moment, it is determined that a fall event has occurred to the target object.

[0050] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a fall protection vest, the fall protection vest comprising:

[0051] A data acquisition module, used to acquire an environmental image and acceleration sensing data; wherein the environmental image and the acceleration sensing data correspond to the motion state of the target object at at least two detection moments;

[0052] The central processing module includes a feature extraction software unit, a feature calculation software unit and a fall determination software unit, wherein the feature extraction software unit is used to extract line features for the environment image at each detection moment to obtain corresponding pattern line features; the feature calculation software unit is used to calculate the rotation angular velocity of the target object according to the pattern line features corresponding to at least two detection moments; the fall determination software unit is used to determine whether the target object falls based on the rotation angular velocity and the acceleration sensor data;

[0053] The fall processing module is used to protect the target object when a fall event occurs to the target object.

[0054] In some embodiments, the data acquisition module includes a first camera device and a second camera device, the first camera device is arranged on the front side of the fall protection vest, and the second camera device is arranged on the shoulder side of the fall protection vest; wherein, the fall protection vest includes block airbags; the block airbags include head and neck airbags, back airbags, left hip and hip airbags, and right hip and hip airbags.

[0055] To achieve the above-mentioned purpose, a third aspect of an embodiment of the present application proposes an electronic device, comprising a memory and a processor, wherein the memory stores a software program, and when the processor executes the software program, the fall detection method described in the first aspect is implemented.

[0056] The fall detection method, fall protection vest and electronic device proposed in the present application obtain environmental images and acceleration sensor data, wherein the environmental images and acceleration sensor data correspond to the motion state of the target object at at least two detection moments, extract line features for the environmental images at each detection moment, obtain corresponding pattern line features, and then calculate the rotational angular velocity of the target object based on the pattern line features corresponding to at least two detection moments, and finally determine whether the target object falls based on the rotational angular velocity and acceleration sensor data. It can be seen from this that the present application calculates the rotational angular velocity of the target object based on the pattern line features at at least two detection moments, thereby accurately obtaining the angular velocity motion information of the target object, and determines whether the target object falls based on the calculated rotational angular velocity and acceleration sensor data, thereby reducing the probability of daily activities being misjudged as falls and improving the accuracy of fall detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of a fall detection method provided in an embodiment of the present application;

[0058] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0059] Figure 3 yes Figure 1 Another flow chart of step S102 in FIG. 1 ;

[0060] Figure 4 yes Figure 3 Another flow chart of step S303 in FIG.

[0061] Figure 5 yes Figure 1 Another flow chart of step S102 in FIG. 1 ;

[0062] Figure 6 yes Figure 1 Another flow chart of step S102 in FIG. 1 ;

[0063] Figure 7 yes Figure 1 Flow chart of step S103 in FIG.

[0064] Figure 8 yes Figure 1 The flowchart after step S102 and step S104 in FIG.

[0065] Fig. 9 is a schematic diagram of the structure of a fall protection vest provided in an embodiment of the present application;

[0066] Fig.10 is a schematic diagram of a fall protection vest provided in an embodiment of the present application;

[0067] Fig.11 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0069] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0070] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0071] The embodiments of the present application provide a fall detection method, a fall protection vest and an electronic device, aiming to improve the accuracy of fall detection.

[0072] The fall detection method, fall protection vest and electronic device provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the fall detection method in the embodiments of the present application is described.

[0073] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0074] Figure 1 is an optional flow chart of the fall detection method provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S104.

[0075] Step S101, acquiring an environmental image and acceleration sensing data; wherein the environmental image and acceleration sensing data correspond to the motion state of the target object at at least two detection moments;

[0076] Step S102, extracting line features from the environment image at each detection moment to obtain corresponding pattern line features;

[0077] Step S103, calculating the rotational angular velocity of the target object according to the pattern line features corresponding to at least two detection moments;

[0078] Step S104: determining whether the target object falls based on the rotational angular velocity and acceleration sensor data.

[0079] Steps S101 to S104 shown in the embodiment of the present application are performed by acquiring environmental images and acceleration sensing data, wherein the environmental images and acceleration sensing data correspond to the motion state of the target object at at least two detection moments, extracting line features for the environmental images at each detection moment, obtaining corresponding pattern line features, and then calculating the rotational angular velocity of the target object based on the pattern line features corresponding to at least two detection moments, and finally judging whether the target object has fallen based on the rotational angular velocity and acceleration sensing data. It can be seen that the present application calculates the rotational angular velocity of the target object based on the pattern line features at at least two detection moments, thereby accurately obtaining the angular velocity motion information of the target object, and judging whether the target object has fallen based on the calculated rotational angular velocity and acceleration sensing data, thereby reducing the probability of daily activities being misjudged as falls and improving the accuracy of fall detection.

[0080] In step S101 of some embodiments, acceleration sensing data can be obtained by an inertial measurement unit (IMU); wherein, the environmental image refers to an image of the surrounding environment obtained by an image collection device, which can be obtained by taking pictures by a camera or other device, or by collecting videos by a camera or other device. The detection moment refers to the moment of detecting the motion state of the target object, and the time interval between two detection moments can be reasonably set. It should be noted that in some cases, the shutter time of the camera is not stable, and the film output speed is different, so the time interval of the detection moment cannot be completely set equal to the camera's shooting interval. It can be that when the camera is taking pictures at high speed, the time interval between the photos is calculated according to the shutter time of the camera, and then the photos are selected according to the time interval between the two detection moments. It is also possible to control the camera to take photos according to the time interval between the two detection moments. If the video is collected by a camera or other device, the image frame data can be read from the video stream according to the time interval between the two detection moments.

[0081] In some embodiments, it is understandable that due to different camera devices or different shooting environments, the acquired images may include many complex situations. Therefore, some acquisition-level processing will be performed when acquiring environmental images, so that the subsequent line feature extraction processing can reduce interference caused by other factors such as equipment or environment.

[0082] In some embodiments, considering that different manufacturers of camera equipment are involved, the transmitted data may have two formats, one is the RGB signal format, i.e., the three color signals of red, green and blue, and the other is the YUV signal format, i.e., the color coding signal, including the brightness signal and the chroma (or color difference) signal. In order not to be affected by the signal format, the data in the RGB signal format is grayed, as shown in the following formula (1):

[0083] Gray=0.299R+0.517G+0.114B (1)

[0084] Among them, R is used to represent the red brightness value, G is used to represent the green brightness value, B is used to represent the blue brightness value, and Gray is used to represent the grayscale value of the pixel after grayscale processing, which is equal to the pixel brightness value.

[0085] For data in YUV signal format, the brightness signal can be directly used as the data of the environment image. By graying the data in RGB signal format, the data in the same format as the YUV signal can be obtained as the environment image, thus avoiding the influence caused by the different device signal formats.

[0086] In some embodiments, the brightness of the environment may also affect the quality of the environment image, so when acquiring the image, the average brightness value and brightness variance of the real-time image may be calculated to detect the brightness of the environment.

[0087] Assume that the real-time image adopts a Cartesian coordinate system, the horizontal axis X is the horizontal direction of the image, the origin O is the lower left corner of the image, and the vertical axis Y is the vertical direction of the image. For the convenience of explaining the contents of the embodiments of the present application, the Cartesian coordinate system will continue to be used in subsequent embodiments. The distance between adjacent pixels is a unit length of 1, and the pixel coordinates are (x, y), where the maximum horizontal coordinate is n and the maximum vertical coordinate is k. The calculation formula (2) of the average brightness value is as follows:

[0088]

[0089] Wherein, f(x, y) represents the brightness value of the pixel point (x, y). If the data is in the RGB signal format, f(x, y) is the Gray value calculated by formula (1). If the data is in the YUV signal format, f(x, y) is the value of the brightness signal.

[0090] Then, the brightness variance V can be calculated based on the average brightness value and the brightness value of the pixel point through formula (3): ar , as follows:

[0091]

[0092] After the average brightness value and brightness variance are calculated, they can be compared with the pre-set threshold parameters. If the average brightness value is less than the first brightness parameter threshold Gt, and the brightness variance is less than the second brightness parameter threshold Dt, the image collection device is allowed to start the high dynamic range mode (High Dynamic Range, HDR) or increase the ISO value of the image collection device to increase the brightness, thereby improving the brightness of the environmental image and avoiding the influence of the brightness of the environment. In order to prevent the first brightness parameter threshold Gt from being easily pulled down by a large number of low-brightness pixels when the brightness of pixels in the image is extremely uneven, the second brightness parameter threshold Dt is also used to constrain the brightness variance of each pixel, thereby avoiding the misjudgment of low overall brightness. Among them, the first brightness parameter threshold Gt can be the average brightness threshold, and the second brightness parameter threshold Dt can be the threshold of the brightness variance of each pixel. These two threshold parameters are related to the parameters of the image collection device.

[0093] In steps S102 to S103 of some embodiments, line features are extracted from the environment image to obtain pattern line features of different patterns in the environment image. The pattern line features of the environment image at at least two detection moments are used to calculate the changes in the pattern line features to obtain the rotational angular velocity of the image. The rotational angular velocity of the image is calculated by calculating the pattern line features, so that more accurate angular velocity motion information of the target object can be obtained.

[0094] It should be noted that the related technology uses an inertial measurement unit to obtain angular velocity data. Since the inertial measurement unit is sensitive to vibration, it is easy to cause erroneous fall event judgments due to the daily life activities of the target object. Image calculation is insensitive to high-frequency vibrations, and the use of images can reduce the misjudgment of fall events. During the activity of the target object, breathing, raising hands and other actions will affect the inertial measurement unit. By obtaining images through the camera, the changes in the environment relative to the person can be obtained, which can reduce the impact of the target object due to daily life activities, thereby reducing the error in the judgment of fall events.

[0095] On the other hand, temperature can easily affect the inertial measurement unit, and large errors will occur when there is a large temperature difference between winter and summer, or between day and night. In the case of large errors, it is also difficult to set the threshold for triggering a fall event. If the threshold is too low, it is easy to accidentally trigger the airbag, which not only wastes the gas cylinder but also scares the target object. If the threshold is too high, the airbag protection may not be triggered when the target object actually falls, causing the target object to be injured by the fall.

[0096] In step S104 of some embodiments, it is determined whether the target object has fallen based on the rotational angular velocity and acceleration sensing data. In S103, the rotational angular velocity of the target object is calculated based on the pattern line features corresponding to at least two detection moments, wherein the two detection moments are generally the previous detection moment and the current detection moment. When the rotational angular velocity exceeds the preset angular velocity threshold, and the acceleration sensing data obtained by the inertial measurement unit at the current detection moment also exceeds the preset acceleration threshold, it is determined that the target object has a fall event at the current detection moment. After determining that a fall event has occurred, protective measures can be taken, such as a fall protection vest opening an airbag to protect the target object.

[0097] In some embodiments, when the target object falls, a large acceleration in the direction perpendicular to the ground is generated, so a more accurate determination of the target object's fall event can be made based on the direction of the acceleration. When determining that a fall event has occurred at the current detection moment, the fall direction of the target user can be determined based on the rotation angular velocity and the positive and negative values ​​of the acceleration sensor, thereby making more accurate protective measures.

[0098] In some embodiments, the fall protection vest is provided with airbags of different blocks, which can open the airbags of different blocks to protect the target object when the target object falls in different postures. Thus, a single large airbag can be divided into multiple small airbags, and a single large air cylinder can also be divided into multiple small air cylinders. This is not only convenient for carrying and replacing air cylinders, but also improves the comfort of use and wear, and can also save the cost of use. In some embodiments, the fall direction of the target user can be determined by combining the positive and negative values ​​of the rotational angular velocity and the acceleration sensor, so as to open the airbags of different blocks, so as to accurately protect the target object, avoid wasting air cylinders, and avoid scaring the target object with excessive noise.

[0099] In some embodiments, multiple image acquisition devices may be provided to simultaneously acquire environmental images, thereby integrating the environmental images obtained by the multiple image acquisition devices to comprehensively determine the fall situation of the target object, including the direction of the fall.

[0100] In some embodiments, after detecting that the elderly have fallen, even with the protection of the airbag, the elderly may still be injured. In order to avoid the elderly from falling into an anxious situation, some pre-recorded voices can be played to remind the elderly to take self-protection measures, or voice help can be sent to people around them. It is also possible to play soft and soothing music to relieve the emotions of the elderly. It is also possible to call the police for help after the fall. The specific time for help can be determined based on the environmental image collected by the image acquisition device. It is also possible to set up a sound acquisition device and comprehensively judge the fall of the elderly by various information. Timely alarm for help can avoid subsequent injuries caused by the fall of the elderly. It can also be sent to designated contacts via text messages to remind relatives to pay attention to the fall of the elderly in a timely manner. Those skilled in the art can achieve foreseeable mitigation measures through a certain degree of free combination.

[0101] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S204:

[0102] Step S201, denoising the environment image to obtain a pure image;

[0103] Step S202, edge detection is performed on the clean image, and non-maximum suppression processing is performed to obtain an edge binary image;

[0104] Step S203, performing edge cutting on the edge binary image based on a predetermined cutting parameter to obtain a preprocessed image;

[0105] Step S204: extract line features from the preprocessed image to obtain pattern line features.

[0106] In step S201 of some embodiments, salt and pepper noise can easily affect the result of line feature extraction in step S102, so the environment image should be denoised first to obtain a pure image. In some embodiments, a median filter is used to filter the salt and pepper noise. The method is to use a (2Na+1)×(2Na+1) convolution kernel to perform a convolution operation on the area to be processed of the environment image, where Na is a positive integer, that is, for a (2Na+1)×(2Na+1) area, the grayscale values ​​are arranged from small to large, and the grayscale value is the first. The gray value of replaces the gray value of the center of the area (2Na+1)×(2Na+1) of the original image. [] represents the floor rounding function, that is, [VI] represents the maximum integer not greater than VI.

[0107] In some embodiments, the environmental image is further subjected to contrast enhancement processing, which may be performed in a linear contrast enhancement manner. The specific calculation formula (4) is as follows:

[0108] g(x,y)=a×f(x,y)+b(4)

[0109] Wherein, a and b are pre-set parameters, and a>1, f(x, y) can represent the brightness value of the pixel point (x, y), and g(x, y) is the brightness value of the pixel point after enhancement.

[0110] In step S202 of some embodiments, the Canny operator is used for edge detection, and the Sobel operator is used to calculate the gradient value and gradient direction. The Sobel operator convolution kernel is 3*3. Then non-maximum suppression (NMS) is performed to remove non-maximum values ​​in the pure image. It should be noted that the purpose of removing non-maximum values ​​is not to obtain the edges in the image, but to avoid retaining too many edge points, which refer to the pixels in the edge binary image. Too many edge points will produce too much noise for subsequent line feature extraction. Too many edge points may be very dense in the Hough space during line feature extraction, which may cause them to be mistakenly voted as an elliptical pattern, and the density of edge points may also cause the voting value to be very unstable, thus generating large noise. On the other hand, due to the high time complexity of elliptical pattern detection, too many edge points will also have a greater impact on the calculation speed.

[0111] In some embodiments, edge points are filtered by dual threshold detection, and the dual thresholds are respectively an upper threshold and a lower threshold. Points with gradient values ​​lower than the lower threshold are determined as non-edge points, and points with gradient values ​​higher than the upper threshold are determined as edge points. For points between the lower threshold and the upper threshold, it is determined whether there are edge points in the 8 adjacent points. If there are, this point is determined as an edge point, and if not, this point is determined as a non-edge point. The gray values ​​of all non-edge points obtained are set to 0, and the gray values ​​of all edge points are set to 255 to obtain an edge binary map. Those skilled in the art can reasonably set the upper threshold and the lower threshold to control the number of edge points in the edge binary map. Too many and too dense edge points will cause large noise when extracting line features, and the voting value is very unstable, and the calculation time required will be a lot, affecting the calculation speed. Too few edge points will be unfavorable for subsequent calculation of the rotation angular velocity.

[0112] In step S203 of some embodiments, the predetermined cutting parameter refers to a ratio parameter for cutting the edge binary image. After obtaining the edge binary image, the edge portion of the edge binary image is cut off. This is because the environmental image collected by the image collection device may have edge distortion and distortion, so not selecting the edge points of the image edge for calculation can also avoid adverse effects caused by the device.

[0113] In some embodiments, 95.45% of the pixel points in the edge binary image in the x and y directions close to the center of the image are selected, and the remaining edge portion of the edge binary image is cut off to obtain a preprocessed image used in subsequent calculations.

[0114] Through steps S201 to S204, the present application performs denoising on the environmental image to obtain a pure image with noise points eliminated, improves the quality of pattern line features obtained by line feature extraction, makes the edges in the pure image more obvious through edge detection, and controls the number of edge points through non-maximum suppression processing, thereby avoiding the generation of large noise during line feature extraction due to the number of edge points being too dense, avoiding the problem of unstable voting values, reducing the time required for calculation, and speeding up the subsequent line feature extraction. By performing edge cutting processing on the edge binary image using predetermined cutting parameters, a preprocessed image is obtained, eliminating the problem of edge distortion and distortion of the environmental image during acquisition due to equipment problems.

[0115] See also Figure 3 In some embodiments, the pattern line feature includes ellipse feature data. Step S102 may include but is not limited to steps S301 to S304:

[0116] Step S301, determining a first number of edge points contained in an environment image;

[0117] Step S302, for the first number of edge points, establishing edge point pairs according to any two edge points to obtain a second number of edge point pairs;

[0118] Step S303, generating a target ellipse pattern for each edge point pair, and determining an ellipse voting value corresponding to the target ellipse pattern;

[0119] Step S304, obtaining corresponding ellipse feature data based on the target ellipse pattern and the ellipse voting value corresponding to the target ellipse pattern.

[0120] In some embodiments, in step S301 to step S302, the first number of edge points contained in the environment image is determined. It can be understood that other image processing can be performed on the environment image before line feature extraction is performed on the environment image. In some embodiments, only denoising and edge detection processing can be performed on the environment image to enable line feature extraction to obtain better pattern line features. For the first number of edge points, edge point pairs are established based on any two edge points to obtain a second number of edge point pairs.

[0121] In step S303 of some embodiments, each edge point pair is processed to generate a target ellipse pattern, and an ellipse voting value corresponding to the target ellipse pattern is determined.

[0122] See also Figure 4 In some embodiments, step S303 may include but is not limited to steps S401 to S405:

[0123] Step S401, determining the major axis endpoints of the ellipse pattern according to the edge point pairs;

[0124] Step S402, determining edge points other than edge point pairs among the first number of edge points as third edge points;

[0125] Step S403, determining a plurality of candidate ellipse patterns according to the edge point pair and each third edge point, and calculating a candidate minor axis length of each candidate ellipse pattern;

[0126] Step S404, determining the ellipse voting value of the corresponding candidate ellipse pattern according to the number of occurrences of each candidate minor axis length;

[0127] Step S405 : determining the candidate ellipse pattern whose ellipse voting value meets the first preset condition as the target ellipse pattern.

[0128] In some embodiments, in step S401 to step S405, a first edge point (t x ,t y ) and the second edge point (u x ,u y ), the first edge point (t x ,t y ) and the second edge point (u x ,u y ) is determined as the major axis endpoint of the elliptical pattern, and a non-first edge point (t x ,t y ) and the second edge point (u x ,u y )'s third edge point (m x ,m y ), according to the first edge point (t x ,t y ), the second edge point (u x ,u y ) and every third edge point (m x ,m y ), and calculate the following formulas (5)-(11):

[0129]

[0130] Among them, x , o y represents the first edge point (t x ,t y ) and the second edge point (ux ,u y ) center point; θ1 represents the inclination angle of the major axis of the elliptical pattern relative to the x-axis of the Cartesian coordinate system; δ represents the third edge point (m x ,m y ) to the center of the ellipse pattern; γ represents the distance from the center of the ellipse pattern to the third edge point (m x ,m y )’s projection length on the major axis of the elliptical pattern; s represents the alternative minor axis length of the alternative elliptical pattern.

[0131] According to the above calculation formulas (5)-(11), each third edge point (m x ,m y ) corresponds to the alternative minor axis length of the alternative elliptical pattern. According to the number of occurrences of each alternative minor axis length, the ellipse voting value of each alternative ellipse pattern is determined. Cumulative voting processing is performed on the alternative minor axis lengths, including calculating the number of minor axes that can be regarded as equivalent within the length error value for each alternative minor axis length, as the ellipse voting value of the corresponding alternative ellipse pattern. Or a length error value is set, the alternative minor axis lengths are divided into different intervals according to the length error value, and the number of alternative minor axis lengths in different intervals is used as the ellipse voting value. It can be understood by those skilled in the art that setting the length error value is to regard approximately equal alternative ellipse patterns as the same ellipse pattern, thereby eliminating the voting value of what should essentially be the same ellipse pattern due to the slightest error in the alternative minor axis length. The cumulative voting processing of the alternative minor axis lengths can also be carried out by the cumulative voting processing method used in the Hough transform.

[0132] It should be noted that the processing method of the ellipse pattern in this embodiment is different from the conventional Hough transform method. When the ellipse pattern is processed by the conventional Hough transform method, the edge point data will be mapped in the five-dimensional Hough space to obtain the ellipse major axis, the ellipse semi-major axis, the ellipse major axis inclination angle, the ellipse center horizontal coordinate and the ellipse center vertical coordinate, and then determine the unique ellipse in the Cartesian coordinate system based on the five-dimensional information. However, the algorithm complexity of the conventional Hough transform for processing ellipse patterns is very high and it will consume a lot of time. Therefore, this embodiment makes an improvement by using the Hough transform of the one-dimensional parameter space to reduce the amount of calculation.

[0133] When the ellipse voting value is greater than the preset first voting threshold, it is considered that the ellipse pattern corresponding to the ellipse voting value is definitely present, and has enough edge points on the ellipse pattern, and has sufficient credibility. The corresponding ellipse pattern is used as the target ellipse pattern. Those skilled in the art can reasonably set the first voting threshold according to the number of edge points in the environmental image. The target ellipse pattern indicates that there is a certain ellipse pattern composed of any edge points in the environmental image. In some embodiments, if the multiple ellipse voting values ​​corresponding to different minor axis lengths in the alternative ellipse patterns obtained based on the first edge point and the second edge point are greater than the preset first voting threshold, then only the ellipse pattern corresponding to the largest ellipse voting value among all ellipse voting values ​​is taken as the target ellipse pattern. It can be understood that among the alternative ellipse patterns obtained by the determined first edge point and the second edge point, only the alternative ellipse pattern corresponding to the largest ellipse voting value in the first voting threshold is selected as the target ellipse pattern, so that the target ellipse pattern obtained is the most reliable, and at the same time, the number of target ellipse patterns is controlled not to be too much, affecting the calculation efficiency.

[0134] When the ellipse voting value meets the first preset condition and exceeds the preset first voting threshold, the corresponding candidate ellipse pattern is determined as the target ellipse pattern. By screening the ellipse voting value according to the first preset condition, the candidate ellipse patterns with fewer edge points and less obvious ellipse features can be filtered out, thereby reducing the interference of some errors, and also making the candidate ellipse patterns with more edge points, higher ellipse voting values, and more obvious ellipse features can be used as target ellipse patterns and used to calculate the ellipse feature data. The credibility of the ellipse feature data is improved, and the accuracy of calculating the rotational angular velocity of the target object based on the pattern line features is improved.

[0135] In step S304 of some embodiments, the ellipse feature data β1 is calculated based on all target ellipse patterns and the ellipse voting values ​​corresponding to the target ellipse patterns using the following formula (12):

[0136]

[0137] Among them, θ1 represents the inclination angle of the major axis of the ellipse pattern relative to the x-axis of the Cartesian coordinate system; z represents the ellipse voting value of the target ellipse pattern. The ellipse feature data is used to represent the angular features of all target ellipse patterns in the environment image, in radians, and is used as a parameter for calculating the rotational angular velocity of the target object later.

[0138] Through step S301 to step S304, by establishing edge point pairs for the first number of edge points, generating a target ellipse pattern based on the edge point pairs, and determining the ellipse voting value corresponding to the target ellipse pattern, compared with the conventional Hough transform ellipse processing, the amount of calculation is reduced, the calculation efficiency is accelerated, and the ellipse feature data can be obtained more quickly. Multiple candidate ellipse patterns are obtained by calculation, and the ellipse voting value is screened by the first preset condition to obtain the target ellipse pattern, thereby obtaining a target ellipse pattern with obvious ellipse features, and finally the ellipse feature data is calculated based on the relevant data of the target ellipse pattern and the corresponding first voting threshold. Thereby, the credibility of the ellipse feature data is increased, and more accurate results can be obtained when the rotational angular velocity of the target object is calculated later.

[0139] See also Figure 5 In some embodiments, the pattern line feature includes straight line feature data, and step S102 may include but is not limited to steps S501 to S503:

[0140] Step S501, performing Hough transform processing on the first number of edge points to obtain a plurality of candidate straight line patterns and straight line voting values ​​corresponding to the candidate straight line patterns;

[0141] Step S502, when the straight line voting value meets the second preset condition, the corresponding candidate straight line pattern is determined as the target straight line pattern;

[0142] Step S503, obtaining corresponding straight line feature data based on the target straight line pattern and the corresponding straight line voting value.

[0143] In steps S501 to S503 of some embodiments, the first number of edge points are subjected to Hough transform, and any edge point (x0, y0) in the environment image is mapped to the ρ-θ two-dimensional parameter space to obtain the expression (13) of the edge point:

[0144] ρ=y0sinθ+x0cosθ (13)

[0145] When there are two curves in the ρ-θ parameter space with an intersection point (ρ0, φ), we can obtain formulas (14) and (15):

[0146] ρ0=y1sinφ+x1cosφ (14)

[0147] ρ0=y2sinφ+x2cosφ (15)

[0148] Combining the above formulas (14) and (15) we get formula (16):

[0149]

[0150] From formula (16), the inclination angle θ2 of the alternative straight line pattern relative to the x-axis is obtained as (17):

[0151] θ2=arctan(-cotφ) (17)

[0152] After mapping all edge points of the environment image into the ρ-θ two-dimensional parameter space, the ρ axis of the ρ-θ space is divided into several small segments, each of which has a length of h (h is a pre-specified constant). The θ axis is then divided into several small segments, each of which has a width of θ0 (θ0 is also a pre-specified constant). This divides ρ-θ into several small squares. Since the length and width of the small squares are relatively small, if s curves converge in this small square, it is considered that there is an alternative straight line pattern within a certain error range, and the straight line voting value of the alternative straight line pattern is s votes.

[0153] When the line voting value satisfies the second preset condition, that is, the line voting value is greater than the preset second voting threshold, it is considered that the candidate line pattern corresponding to the line voting value does exist, and there are enough edge points on the candidate line pattern, which has a certain degree of credibility, and the candidate line pattern is used as the target line pattern. Those skilled in the art can reasonably set the second voting threshold according to the number of edge points in the environment image.

[0154] By filtering the straight line voting values ​​through the second preset condition, it is possible to filter out those candidate straight line patterns with fewer edge points, less obvious straight line features, and fewer edge points, thereby reducing the interference of some error patterns, and also making the candidate straight line patterns with more edge points, higher straight line voting values, and more obvious straight line features as target straight line patterns, and used to calculate the straight line feature data. This improves the credibility of the straight line feature data and the accuracy of calculating the rotational angular velocity of the target object based on the pattern line features.

[0155] Based on all target straight line patterns, the straight line feature data β2 is calculated using the following formula (18):

[0156]

[0157] Among them, θ2 represents the angle value of the target straight line pattern in the two-dimensional parameter space; x represents the straight line voting value of the target straight line pattern. The straight line feature data is used to represent the angle feature of the target straight line pattern in the environment image, in radians, and is used as a parameter for calculating the rotational angular velocity of the target object later.

[0158] Through steps S501 to S503, by performing Hough transform processing on the edge points, multiple candidate straight line patterns and corresponding straight line voting values ​​are obtained, and the straight line voting values ​​of the candidate straight line patterns are screened according to the second preset condition to obtain a target straight line pattern with more obvious straight line features and higher credibility, and the straight line feature data is calculated by the target straight line pattern and the corresponding straight line voting value, thereby obtaining straight line feature data with high credibility, so as to obtain more accurate results when calculating the rotational angular velocity of the target object later.

[0159] See also Figure 6 In some embodiments, the pattern line feature includes circular feature data, and step S102 may also include but is not limited to steps S601 to S607:

[0160] Step S601, selecting two environment images at two detection moments from the environment image at each detection moment; wherein the two detection moments are respectively a first detection moment and a second detection moment;

[0161] Step S602, performing Hough transform processing on the edge points at the first detection moment to obtain a plurality of first candidate circular patterns and corresponding first circular voting values;

[0162] Step S603, performing Hough transform processing on the edge points at the second detection moment to obtain a plurality of second candidate circular patterns and corresponding second circular voting values;

[0163] Step S604, when the first circular voting value and the second circular voting value meet the third preset condition, and the relative circle center distance meets the fourth preset condition, the first target circular pattern at the first detection moment and the second target circular pattern at the second detection moment are obtained; wherein the relative circle center distance represents the distance between the center of the first candidate circular pattern corresponding to the first circular voting value and the center of the second candidate circular pattern corresponding to the second circular voting value;

[0164] Step S605, obtaining a circle center change angle based on the first target circular pattern and the second target circular pattern;

[0165] Step S606, obtaining a circular pattern voting mean value according to a first circular voting value corresponding to the first target circular pattern and a second circular voting value corresponding to the second target circular pattern;

[0166] Step S607, obtaining circular feature data based on the circle center change angle, the circular pattern voting mean, the first target circular pattern and the second target circular pattern.

[0167] In steps S601 to S607 of some embodiments, two environment images at each detection moment are selected from the environment images at each detection moment and are used as the first detection moment and the second detection moment respectively;

[0168] The edge points in the environment image at the first detection moment are subjected to Hough transformation to obtain the first candidate circular pattern in the three-dimensional parameter space (a, b, r), where a and b represent the coordinates of the center of the circular pattern, and r represents the radius of the circular pattern. Cumulative voting is performed on multiple first candidate circular patterns to obtain the first circular voting value of the corresponding circular pattern. Then, the edge points in the environment image at the second detection moment are subjected to Hough transformation to obtain the second candidate circular pattern in the three-dimensional parameter space, and cumulative voting is performed on multiple second candidate circular patterns to obtain the second circular voting value of the corresponding circular pattern.

[0169] When the first circular voting value and the second circular voting value meet the third preset condition, that is, when both the first circular voting value and the second circular voting value are greater than the third voting threshold, and when the relative center distance meets the fourth preset condition, that is, the relative center distance is less than the center distance threshold, the first candidate circular pattern corresponding to the first circular voting value is used as the first target circular pattern, and the second candidate circular pattern corresponding to the second circular voting value is used as the second target circular pattern. The fourth preset condition can also add the condition of the circle radius, that is, when the relative center distance is less than the center distance threshold, the circle radius of the first candidate circular pattern and the second candidate circular pattern are equal within a certain error.

[0170] Among them, the relative center distance is to calculate the center distance between the first candidate circular pattern and the second candidate circular pattern. By setting the center distance threshold, it can be limited that the candidate circular patterns at the first detection moment and the second detection moment are essentially the same circular pattern, so that the displacement information of the same circular pattern at different detection moments can be obtained. Afterwards, the center change angle and circular feature data of the same circular pattern at different detection moments are calculated, so that the calculation result obtained can reflect the motion information of the target object within the two detection moments. Therefore, it can be understood that when the first circular voting value and the second circular voting value meet the third preset condition and the relative center distance meets the screening of the fourth preset condition, the corresponding first candidate circular pattern and the second candidate circular pattern will no longer participate in the subsequent screening after being determined to be the same circular pattern. That is, after the first candidate circular pattern determines the corresponding second candidate circular pattern under the constraints of the third preset condition and the fourth preset condition, it can be considered that they are the same circular pattern at different detection moments, so in the subsequent processing of the candidate circular pattern, the candidate circular pattern at different detection moments that has determined the same circular pattern will no longer appear, avoiding meaningless data.

[0171] Based on the first target circular pattern and the second target circular pattern, the circle center change angle is calculated. The specific calculation formula (19) is as follows:

[0172]

[0173] Among them, a i and b i A represents the coordinate voting value of the center of the first target circular pattern; i and B i Represents the coordinate voting value of the center of the second target circular pattern; θ 3i The center change angle of the first target circular pattern and the second target circular pattern can be calculated by formula (19) to obtain the center change angle of the same circular pattern at two detection moments. The center change angle represents the difference in the inclination angle between the center of the same circular pattern and the origin of the Cartesian coordinate system at two detection moments.

[0174] Then, the circular pattern voting mean is obtained according to the first circular voting value corresponding to the first target circular pattern and the second circular voting value corresponding to the second target circular pattern.

[0175] Finally, based on the circle center change angle, the circular pattern voting mean, the first target circular pattern and the second target circular pattern, the circular feature data is calculated. The specific calculation formula (20) is as follows:

[0176]

[0177] Among them, r i Indicates the radius value of the first target circular pattern, R i Indicates the radius value of the second target circular pattern; X i The average value of the total voting values ​​of the first target circular pattern and the second target circular pattern. The circular feature data represents the angle feature change value of the same circular pattern obtained by combining all the first target circular pattern and the second target circular pattern within two detection moments.

[0178] It should be noted that the circular feature data obtained here is not the same as the elliptical feature data mentioned in the previous step S304 and the straight line feature data mentioned in step S503. The elliptical feature data and the straight line feature data are both feature data of the pattern at a certain detection time, but the circular feature data is the feature change value of the circular pattern between two detection times. This is because the linear pattern and the elliptical pattern can directly obtain information at the rotation angle level, but the circular pattern cannot directly obtain information at the rotation angle level from the circular pattern due to its own rotation invariance. Therefore, it is necessary to determine the information of the circular pattern at the rotation angle level based on the feature change value of the same circular pattern at two times.

[0179] In addition, the elliptical pattern has a higher complexity than the circular pattern, and the long axis inclination angle of the elliptical pattern is a significant and difficult to change feature. Therefore, when setting the first voting threshold and the third voting threshold, they need to be set higher than the second voting threshold, so that the target elliptical pattern and the first target circular pattern and the second target circular pattern obtained by voting have higher credibility. It should be noted that the target object's usual turning action or daily activities have little effect on the pattern line features, because the environmental image does not have much rotational movement, but more of the environmental image is translated. However, when the body tilts or falls, the information at the rotation angle level of these patterns will be more obvious. If the target object only lies down, the pattern line features may be slightly larger, but even if a slightly larger rotation angular velocity is calculated, the acceleration sensor data generated is not large because the elderly move more slowly and more carefully. Therefore, based on the rotation angular velocity and acceleration sensor data, it will not easily misjudge the fall event.

[0180] It should be noted that in the calculation formula (12) of the ellipse feature data, the calculation formula (18) of the straight line feature data, and the calculation formula (20) of the circular feature data, the role of the denominator is normalization, ensuring that the calculation result is still the angle value, and ensuring that the voting value cannot affect the size of the angle value, but can only affect the weight of the angle value of different patterns in the calculation process, or can be understood as the influence of different patterns on the calculation result. The calculation process of the angle value can be intuitively understood as the dot product of the angle value vector and the voting value vector. Taking the ellipse pattern as an example, both vectors are i-dimensional, and the i-th dimension represents the major axis inclination angle of the i-th ellipse pattern that exceeds the voting value and the corresponding ellipse voting value, and the voting value vector needs to be normalized and then dot product, and the denominator is used to normalize the voting value vector. Therefore, when the voting values ​​of all target ellipse patterns are scaled up, the angle finally calculated will not change, that is, the voting value cannot affect the size of the angle value. The target ellipse pattern with a higher voting value will be assigned a greater influence. For example, when the voting value of the target ellipse pattern with a large angle value is high and the voting value of the target ellipse pattern with a small angle value is low, the feature data finally calculated will be relatively larger. It should be noted that the actual purpose of this process of calculating feature data is to integrate various information in the image and the corresponding reliability, so as to eliminate noise interference and obtain a robust result. Taking the ellipse as an example, every long-axis inclination angle of the target ellipse pattern that exceeds the first voting threshold can participate in the calculation of the ellipse feature data, but the higher the ellipse voting value, the higher the reliability of the target ellipse pattern.

[0181] The calculation formula (20) for circular feature data uses the Gaussian function to reduce the impact of circular patterns with large radius differences on the calculation results. The parameter is used 5 times, thereby reducing the smoothing area of ​​the Gaussian function close to 0, so that even if r i -R i The numerical value of is not large, but the two circular patterns at the first moment and the second moment do not belong to the same circle, which basically will not have much impact on the calculated circular feature data. In this case, the result of the target circular pattern calculation that the two circular patterns at the first moment and the second moment are the same circle will also be more affected. Formula (20) uses the exponential form of e, and its main purpose is to eliminate the impact caused by the circular patterns at the first moment and the second moment not being the same circle. Although the relative center distance is calculated before calculating the center change angle, and the center distance threshold is used for screening, such a situation may still occur. By The part can reduce the impact caused by the circular pattern at the first moment and the circular pattern at the second moment being different from each other. The part can be understood as the average voting value of the same circular pattern at the first moment and the second moment. The design idea is to let the circular pattern with a higher average voting value, that is, the circular pattern with a higher credibility, influence the calculation of the circular feature data with a greater weight. Similarly, in the calculation formula (12) for the elliptical feature data, the cube of the elliptical voting value is set to further amplify the influence of the target elliptical pattern with a high elliptical voting value. Even one or several target elliptical patterns with very high elliptical voting values ​​can completely determine the calculation results of the elliptical feature data. For the calculation formula (18) for the straight line feature data, it is set to the natural logarithm of the straight line voting value. Compared with the simple weighted average of the voting values, it can slightly highlight the influence of the target straight line pattern with a high straight line voting value on the calculation of the straight line feature data, but not too much.

[0182] Through steps S601 to S607, by performing Hough transform processing on the edge points at two detection moments, and by setting the third preset condition and the fourth preset condition, the first target circular pattern and the second target circular pattern are obtained, thereby ensuring as much as possible that any first target circular pattern has a corresponding second target circular pattern, the first target circular pattern and the second target circular pattern are essentially the same circular pattern, and both have a certain degree of credibility, and the displacement information of the same circular pattern at different detection moments can be calculated based on the first target circular pattern and the second target circular pattern, so that the circular feature data can be accurately obtained, reflecting the motion information of the target object at the two detection moments.

[0183] See also Figure 7In some embodiments, two of the detection moments are respectively used as the first target moment and the second target moment, and step S103 includes but is not limited to steps S701 to S703:

[0184] Step S701, obtaining first deflection data according to the ellipse feature data and the straight line feature data at the first target moment;

[0185] Step S702, obtaining second deflection data according to the ellipse feature data and the straight line feature data at the second target moment;

[0186] Step S703: Calculate the rotation angular velocity of the target object based on the first deflection data, the second deflection data and the circular feature data.

[0187] In steps S701 to S703 of some embodiments, the corresponding first deflection data αt1 is obtained by adding the elliptical feature data β1 and the straight line feature data β2 at the first target time t1, and the corresponding second deflection data αt2 is obtained by adding the elliptical feature data β1 and the straight line feature data β2 at the second target time t2, and then the rotational angular velocity of the target object is calculated based on the first deflection data, the second deflection data and the circular feature data β3. The specific calculation formula (21) is as follows:

[0188]

[0189] Wherein, w represents the rotational angular velocity of the target object. The ellipse feature data and the straight line feature data are both angular features at a certain detection time, so they need to be calculated based on the time change, while the circular feature data is the angular feature change value obtained based on two different detection times.

[0190] Through steps S701 to S703, the elliptical feature data and the straight line feature data at the first target moment and the second target moment, as well as the circular feature data at the first target moment and the second target moment, are combined with the feature data of the three patterns to calculate the rotational angular velocity of the target object, thereby determining the angular velocity motion information of the target object. Compared with the related art of obtaining the angular velocity motion information of the target through an inertial measurement unit, the error caused by the vibration generated by the movement of the target object can be reduced, and more accurate angular velocity motion information can be obtained.

[0191] In some embodiments, when the target object falls, a large acceleration in the direction perpendicular to the ground is generated, so a more accurate determination of the target object's fall event can be made based on the direction of the acceleration. When determining that a fall event has occurred at the current detection moment, the fall direction of the target user can be determined based on the rotation angular velocity and the positive and negative values ​​of the acceleration sensor, thereby making more accurate protective measures.

[0192] See also Figure 8 In some embodiments, step S102 may include but is not limited to steps S801 to S804:

[0193] Step S801, recording the number of target elliptical patterns at the detection moment as a first determination number;

[0194] Step S802, recording the number of target straight line patterns at the detection time as a second determination number;

[0195] Step S803, recording the number of the first target circular pattern or the second target circular pattern at the detection time as a third determination number;

[0196] Step S804, calculating the image feature value based on the first determination number, the second determination number and the third determination number to obtain the image feature value at the detection time;

[0197] Step S104 may include but is not limited to steps S805 to S806:

[0198] Step S805, based on the image feature value at each detection moment, obtaining a quantity feature change value in a preset time series; wherein the preset time series is composed of a third number of detection moments;

[0199] Step S806, when the quantity characteristic change value satisfies the fifth preset condition at the third target moment in the preset time sequence, and the quantity characteristic change value satisfies the sixth preset condition after the third target moment, it is determined that a fall event has occurred to the target object;

[0200] In some embodiments, in steps S801 to S804, when extracting line features based on an environment image at a certain detection time t, when the ellipse voting value is greater than the first voting threshold, the corresponding ellipse pattern is used as a target ellipse pattern, and the number of target ellipse patterns is recorded as a first determination number. Similarly, the number of target straight line patterns is recorded as a second determination number, and the number of first target circular patterns or second target circular patterns is recorded as a third determination number.

[0201] Based on the first determination number, the second determination number and the third determination number, the image feature value is calculated to obtain the image feature value at the detection time. The specific formula (22) is as follows:

[0202] f(t)=k1o t 2 +k2c t 2 +k3l t (twenty two)

[0203] Among them,t represents the first judgment quantity, l t represents the second judgment quantity, c t represents the third determination quantity, and k1, k2, and k3 are preset weight coefficients.

[0204] The image feature value is used to reflect the quantitative features of the target elliptical pattern, the target straight line pattern and the first target circular pattern (or the second target circular pattern) within a certain detection moment.

[0205] After determining whether the target object has fallen based on the rotational angular velocity and acceleration sensor data, through steps S805 to S806, the absolute value of the first-order difference value on the preset time series is obtained according to the image feature values ​​at multiple consecutive detection moments, thereby obtaining the quantitative feature change value g(t i ), the calculation formula (23) is as follows:

[0206] g(t i )=|f(t i )-f(t i-1 )| (23)

[0207] Among them, the preset time series is composed of a third number of detection moments, wherein the third number can be reasonably set by a technician in this field. When the quantity feature change value satisfies the fifth preset condition at the third target moment on the preset time series, and the quantity feature change value after the third target moment satisfies the sixth preset condition, that is, the quantity feature change value is greater than the first change threshold, and the quantity feature change values ​​at the fourth number of consecutive detection moments after the third target moment are less than the second change threshold, it is determined that a fall event has occurred in the target object. It should be noted that the quantity feature change value can reflect the change of the image feature value, and then reflect the change of the environmental image at different detection moments.

[0208] No device can guarantee absolute accuracy in detecting elderly people falling. If the target object falls but the airbag does not open, in order to rescue the target object in time, the change value of the quantity feature calculated over a period of time can be used to determine whether a fall has occurred. When the elderly person falls, the environmental image obtained by the image acquisition device will change dramatically, thereby affecting the number of target elliptical patterns, target straight line patterns and first target circular patterns in the environmental image. Experience shows that when the elderly person falls, the number of target elliptical patterns and first target circular patterns in the environmental image will be greatly reduced, and may even be zero, while the number of straight lines changes less. Therefore, in formula (22), the first judgment number and the third judgment number are set in square form. This is to make the change in the number of target elliptical patterns and first target circular patterns in the environmental image more obvious after the elderly person falls, so that it is clearly reflected in the change value of the quantity feature. Therefore, when the quantity feature change value at the third target moment is greater than the first change threshold, it means that a large change has occurred in the environmental image at the third target moment, and a fall event may have occurred. After the third target moment, the quantity feature change values ​​at the fourth number of consecutive detection moments are all less than the second change threshold, which means that the change in the environmental image in the environmental image is small. It may be that after a fall, the target object is no longer active on the ground, or the activity is small.

[0209] In some embodiments, the determination of the change value of the acceleration sensor data can be added to the fifth preset condition, and the change of the acceleration sensor data within a period of time can be added to the sixth preset condition, and the quantitative characteristic change value can be combined to comprehensively determine whether the target object has a fall event at a previous detection moment. It can also be combined with the data of the sound collection device. If the target object falls, but the fall determination is not triggered, a loud sound will be generated when the target object falls. If the target object falls and is determined to have fallen, the expansion of the airbag will also make a loud sound. Therefore, combining the data of the sound collection device is very helpful for re-determining whether a fall event has occurred.

[0210] In some embodiments, by re-determining whether a fall event has occurred, the consequences caused by misjudging the fall event of the target object can be avoided. If the target object has a fall event, the target object can be helped in time by re-determining it, which can also be used as an opportunity to call the police for help.

[0211] If the target object did not fall, but the device mistakenly determined that a fall occurred, re-confirmation can be used to avoid alarming for help and further handling of the fall, thereby reducing resource waste or reducing the adverse effects caused by erroneous judgments.

[0212] Through steps S801 to S806, the quantitative feature change value can be calculated through the image feature values ​​at multiple consecutive detection moments, and the adverse effects of false fall events can be avoided by setting the fifth preset condition and the sixth preset condition.

[0213] See also Fig. 9 and Fig.10 The embodiment of the present application also provides a fall protection vest, which can implement the above-mentioned fall detection method, and the fall protection vest includes:

[0214] A data acquisition module, used to acquire environmental images and acceleration sensing data; wherein the environmental images and acceleration sensing data correspond to the motion state of the target object at at least two detection moments;

[0215] The central processing module includes a feature extraction software unit, a feature calculation software unit and a fall determination software unit, wherein the feature extraction software unit is used to extract line features from the environment image at each detection moment to obtain the corresponding pattern line features; the feature calculation software unit is used to calculate the rotation angular velocity of the target object according to the pattern line features corresponding to at least two detection moments; the fall determination software unit is used to determine whether the target object falls based on the rotation angular velocity and acceleration sensor data;

[0216] The fall processing module is used to protect the target object when a fall event occurs to the target object.

[0217] In some embodiments, the data acquisition module includes a first camera device and a second camera device, wherein the first camera device is arranged on the front side of the fall protection vest, and the second camera device is arranged on the shoulder side of the fall protection vest; wherein the fall protection vest includes block airbags; the block airbags include head and neck airbags, back airbags, left hip airbags and right hip airbags. The fall protection vest is provided with different block airbags, and when the target object falls in different postures, the airbags of different blocks can be opened to protect the target object. Thus, a single large airbag is divided into multiple small airbags, and similarly a single large air cylinder can also be divided into multiple small air cylinders, which is not only convenient to carry and replace the air cylinder, improves the comfort of use and wearing, but also saves the cost of use.

[0218] In an exemplary embodiment, two rotational angular velocities can be acquired by the first camera device and the second camera device. When any one of the rotational angular velocities exceeds a preset angular velocity threshold, and the acceleration sensing data at the current detection moment also exceeds a predetermined acceleration threshold, it is determined that the target object is experiencing a fall event. At this time, the fall posture of the target object can be determined based on the rotational angular velocity and the acceleration sensing data.

[0219] The right hand direction of the target object is set as the positive direction of the x-axis, the front is set as the positive direction of the z-axis, and the vertical downward direction is set as the positive direction of the y-axis. The angular velocity threshold includes a first angular velocity threshold and a second angular velocity threshold.

[0220] When the target object falls, the first camera device is arranged on the front side of the fall protection vest, including the chest of the target object, to obtain a first rotation angular velocity. The second camera device is arranged on the shoulder side of the fall protection vest, including the left shoulder and the right shoulder of the target object, to obtain a second rotation angular velocity.

[0221] If the absolute value of the first rotational angular velocity is greater than the first angular velocity threshold, the first rotational angular velocity direction is negative on the x-axis, and the absolute value of the acceleration sensing data in the x-axis direction is greater than the acceleration threshold, it is determined that the target object is falling to the left.

[0222] If the absolute value of the first rotational angular velocity is greater than the first angular velocity threshold, the first rotational angular velocity direction is positive on the x-axis, and the absolute value of the acceleration sensor data in the x-axis direction is greater than the acceleration threshold, it is determined that the target object is falling to the right.

[0223] If the absolute value of the second rotational angular velocity is greater than the second angular velocity threshold, the second rotational angular velocity direction is positive on the z-axis, and the absolute value of the acceleration sensor data on the z-axis is greater than the acceleration threshold, it is determined that the target object is falling forward.

[0224] If the absolute value of the second rotational angular velocity is greater than the second angular velocity threshold, the second rotational angular velocity direction is negative on the z-axis, and the absolute value of the acceleration sensor data on the z-axis is greater than the acceleration threshold, it is determined that the target object is falling forward.

[0225] It is understandable that if the acceleration sensing data on the y-axis exceeds the acceleration threshold, it can assist in determining whether a fall event has occurred to the target object. Although the y-axis cannot reflect the fall direction of the target object, it is important for determining a fall event. It is understandable that the fall direction of the target object can also be assisted in determining the fall direction based on the positive and negative values ​​of the acceleration sensing data. The above-mentioned multiple acceleration thresholds can set different acceleration thresholds according to the fall direction, so that the fall direction of the target object can be determined more accurately. At this time, the corresponding block airbags can be opened according to the fall direction of the target object to accurately protect the target object. In some embodiments, the acceleration threshold can be set to 0.3g, where g represents the acceleration of gravity.

[0226] In an exemplary embodiment, the fall detection method described in the first aspect of the present application is integrated into a wearable fall protection vest, with a camera on the front and left side of the vest, and providing video data in YUV signal format.

[0227] The camera is connected to the central processing module, which is composed of a CPU, a memory, an acceleration sensor, a gas cylinder control switch, etc. The central processing module controls the opening of the gas cylinder. The gas cylinder is connected to the airbag, which is inside the vest and is not inflated at ordinary times. There are four airbags, each connected to a gas cylinder. The four airbags are the head and neck airbag, the back airbag, the left hip airbag and the right hip airbag.

[0228] The acceleration sensor is a three-axis acceleration sensor. The acceleration sensor is connected to the central processing module. The acceleration values ​​of each axis of the acceleration sensor are input into the central processing module in real time. And the acceleration sensor is connected to the wake-up pin of the CPU of the central processing module. The central processing module is connected to the alarm module, and the alarm module includes a text message sending unit and a voice alarm unit.

[0229] The SMS sending unit has a built-in SIM card slot, and a pre-installed SIM card is configured to complete the sending parameters. The SMS sending unit has a reserved USB interface so that it can be connected to a computer via USB, and the user can edit the recipient, SMS content, etc. through computer software. The SMS sending unit is connected to the central processing module and the antenna. After the central processing module gives a signal to send the SMS, the SMS sending unit sends the SMS to the designated recipient.

[0230] The voice alarm unit has a built-in voice chip, which stores music and alarm voice. The language chip uses an ISD series chip. The voice alarm unit is connected to the central processing module and an 8-ohm speaker. When the central processing module gives an alarm signal, the voice alarm unit emits a preset sound.

[0231] The central processing module is connected to the microphone module. The microphone module is an input module. After the microphone obtains the sound signal, it is first amplified, and then the volume obtained by the microphone is compared with the reference volume level value provided by the central processing module. If it is greater than the reference, the signal is output to the central processing module. The entire system can be powered by batteries.

[0232] The central processing module automatically enters a dormant state after being stationary for 10 minutes. After the user puts on the vest, the acceleration sensor senses the action of putting on the vest and transmits the signal to the central processing module, thereby waking up the central processing module from dormancy. The camera starts working and transmits the video stream to the central processing module. The CPU of the central processing module captures an image from the video stream as the environmental image every △t (0.05s). Determine whether the brightness is too low according to the algorithm described above: Since a camera that provides a YUV signal is used, the data of the brightness signal channel can be used directly. The environmental image is then processed according to the fall detection method described in the first aspect of the present application.

[0233] Among them, in the non-maximum suppression process, the current mainstream and mature Non-Maximum Suppression algorithm is used here. This algorithm is based on the John Canny algorithm to detect the four gradient directions of 0, 45, 90, and 135, and is improved to detect the gradient value in any direction. The method is to generate sub-pixel points based on the gradient direction and the brightness value of the nearby points, and compare the brightness value of the center point with each sub-pixel point to determine whether it is a maximum point.

[0234] In the dual threshold detection part, points with a gradient lower than the lower threshold are judged as non-edge points, and points with a gradient higher than the upper threshold are judged as edge points. For points between the lower threshold and the upper threshold, it is determined whether there are edge points among the 8 adjacent points. If there are, this point is judged as an edge point. If not, this point is judged as a non-edge point. Set the upper threshold TH and the lower threshold TL to a lower level, and take TH:TL=2:1. After completing the dual threshold detection, you can get the edge binary image.

[0235] The use of the central processing module CPU makes it possible to intelligently control multiple airbags in blocks. The block multi-airbag solution can divide the gas cylinder into several parts, which is more comfortable than stuffing a large gas cylinder in a vest.

[0236] The fall protection vest is divided into four airbags, namely the head and neck airbag, which surrounds the head and neck and fully covers the head and neck of the user; the back airbag, which goes up to the end of the cervical spine, the lower edge of the head and neck airbag, which goes down to the fifth joint of the lumbar spine, and the upper edge of the hip airbag, which is the same width as the user's shoulder blades, covering the entire back of the user; the hip airbag is divided into two on the left and right, extending forward to cover the ilium. On the one hand, the protruding ilium is easy to be injured in a fall, and on the other hand, the sacroiliac joint or femoral head of the elderly is relatively fragile and needs protection. The airbag goes down to the farthest end of the sacrum and covers the entire buttocks.

[0237] When the central processing module determines that the user is falling, the gas cylinder is opened, and the high-pressure gas quickly fills the airbag to protect the user. After determining the falling posture, the following are the treatments for various situations: Falling forward: open the head and neck airbags and the hip and hip airbags on both sides. Because there may be a roll after a forward fall, open the hip and hip airbags to protect the ilium and related joints. Falling left or right: open the head and neck airbags and the hip and hip airbags on the side of the fall. Falling backward: all airbags are opened.

[0238] The fall detection method, fall protection vest and electronic device proposed in the present application obtain environmental images and acceleration sensor data, wherein the environmental images and acceleration sensor data correspond to the motion state of the target object at at least two detection moments, extract line features for the environmental images at each detection moment, obtain corresponding pattern line features, and then calculate the rotational angular velocity of the target object based on the pattern line features corresponding to at least two detection moments, and finally determine whether the target object falls based on the rotational angular velocity and acceleration sensor data. It can be seen from this that the present application calculates the rotational angular velocity of the target object based on the pattern line features at at least two detection moments, thereby accurately obtaining the angular velocity motion information of the target object, and determines whether the target object falls based on the calculated rotational angular velocity and acceleration sensor data, thereby reducing the probability of daily activities being misjudged as falls and improving the accuracy of fall detection.

[0239] An embodiment of the present application further provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a software program, and the processor executing the program to implement the above-mentioned fall detection method.

[0240] See also Fig.11 , Fig.11 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0241] The processor 1101 may be implemented by SOC, DSP, or one or more ASIC integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0242] The memory 1102 includes a program memory FLASH Memory and a memory (RAM) required for program operation. The memory 1102 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1102, and the processor 1101 is called to execute the fall detection method of the embodiment of the present application;

[0243] Input / output interface 1103, used to implement information input and output;

[0244] The communication interface 1104 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0245] The processor 1101 , the memory 1102 , the input / output interface 1103 and the communication interface 1104 are connected to each other inside the device.

[0246] An embodiment of the present application further provides a readable and writable storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned fall detection method is implemented.

[0247] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory can include a high-speed random access memory RAM, which is used to store various data when the program is running, and can also include a non-transitory memory, such as a flash memory device FLASH, or other non-transitory solid-state storage devices.

[0248] The fall detection method, fall protection vest and electronic device provided by the embodiment of the present application obtain environmental images and acceleration sensor data, wherein the environmental images and acceleration sensor data correspond to the motion state of the target object at at least two detection moments, extract line features for the environmental images at each detection moment, obtain corresponding pattern line features, and then calculate the rotation angular velocity of the target object according to the pattern line features corresponding to at least two detection moments, and finally determine whether the target object falls based on the rotation angular velocity and acceleration sensor data. It can be seen that the present application calculates the rotation angular velocity of the target object according to the pattern line features at at least two detection moments, thereby accurately obtaining the angular velocity motion information of the target object, and determines whether the target object falls based on the calculated rotation angular velocity and acceleration sensor data, thereby reducing the probability of daily activities being misjudged as falls and improving the accuracy of fall detection. .

[0249] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0250] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0251] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0252] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0253] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0254] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0255] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. 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. 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.

[0256] The units described above 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.

[0257] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0258] 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 readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product.

[0259] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A fall detection method, characterized in that: The method comprises: Acquire an environmental image and acceleration sensing data; wherein the environmental image and the acceleration sensing data correspond to the motion state of the target object at at least two detection moments; Extracting line features from the environment image at each detection moment to obtain corresponding pattern line features; The rotational angular velocity of the target object is calculated according to the pattern line features corresponding to at least two of the detection moments; wherein the pattern line features include elliptical feature data, straight line feature data, and circular feature data; taking two of the detection moments as the first target moment and the second target moment, respectively, the rotational angular velocity of the target object is calculated according to the pattern line features corresponding to at least two of the detection moments, including: obtaining first deflection data according to the elliptical feature data and the straight line feature data at the first target moment; obtaining second deflection data according to the elliptical feature data and the straight line feature data at the second target moment; and calculating the rotational angular velocity of the target object based on the first deflection data, the second deflection data, and the circular feature data; Based on the rotational angular velocity and the acceleration sensing data, it is determined whether the target object falls.

2. The method according to claim 1, characterized in that The extracting line features of the environment image at each detection moment to obtain corresponding pattern line features includes: Performing denoising on the environmental image to obtain a pure image; Performing edge detection on the clean image and performing non-maximum suppression processing to obtain an edge binary image; Performing edge cutting on the edge binary image based on a predetermined cutting parameter to obtain a preprocessed image; Line features are extracted from the preprocessed image to obtain the pattern line features.

3. The method according to claim 1, characterized in that The extracting line features of the environment image at each detection moment to obtain corresponding pattern line features includes: Determining a first number of edge points included in the environment image; For the first number of edge points, establish edge point pairs according to any two of the edge points to obtain a second number of edge point pairs; Generate a target ellipse pattern for each edge point pair, and determine an ellipse voting value corresponding to the target ellipse pattern; Based on the target ellipse pattern and the ellipse voting value corresponding to the target ellipse pattern, the corresponding ellipse feature data is obtained.

4. The method according to claim 3, characterized in that The generating a target ellipse pattern for each edge point pair and determining an ellipse voting value corresponding to the target ellipse pattern includes: Determining the major axis endpoints of the elliptical pattern according to the edge point pairs; Determine the edge points other than the edge point pairs among the first number of edge points as third edge points; Determine a plurality of candidate ellipse patterns according to the edge point pair and each of the third edge points, and calculate a candidate minor axis length of each of the candidate ellipse patterns; Determining the ellipse voting value of the corresponding candidate ellipse pattern according to the number of occurrences of each candidate short axis length; The candidate ellipse pattern whose ellipse voting value meets the first preset condition is determined as the target ellipse pattern.

5. The method according to claim 3, characterized in that: The extracting line features of the environment image at each detection moment to obtain corresponding pattern line features includes: Performing Hough transform processing on the first number of edge points to obtain a plurality of candidate straight line patterns and straight line voting values ​​corresponding to the candidate straight line patterns; When the straight line voting value meets the second preset condition, the corresponding candidate straight line pattern is determined as the target straight line pattern; Based on the target straight line pattern and the corresponding straight line voting value, the corresponding straight line feature data is obtained.

6. The method according to claim 5, characterized in that The extracting line features of the environment image at each detection moment to obtain corresponding pattern line features includes: Selecting two of the environment images at the detection moments from the environment images at each detection moment; wherein the two detection moments are respectively a first detection moment and a second detection moment; Performing Hough transform processing on the edge points at the first detection moment to obtain a plurality of first candidate circular patterns and corresponding first circular voting values; Performing Hough transform processing on the edge points at the second detection moment to obtain a plurality of second candidate circular patterns and corresponding second circular voting values; When the first circular voting value and the second circular voting value meet the third preset condition, and the relative circle center distance meets the fourth preset condition, the first target circular pattern at the first detection moment and the second target circular pattern at the second detection moment are obtained; wherein the relative circle center distance represents the distance between the center of the first candidate circular pattern corresponding to the first circular voting value and the center of the second candidate circular pattern corresponding to the second circular voting value; Based on the first target circular pattern and the second target circular pattern, obtaining a circle center change angle; Obtaining a circular pattern voting mean value according to a first circular voting value corresponding to the first target circular pattern and a second circular voting value corresponding to the second target circular pattern; The circular feature data is obtained based on the circle center variation angle, the circular pattern voting mean, the first target circular pattern and the second target circular pattern.

7. The method according to claim 6, characterized in that The extracting line features of the environment image at each detection moment to obtain corresponding pattern line features also includes: Recording the number of the target elliptical patterns at the detection moment as a first determination number; recording the number of the target straight line patterns at the detection moment as a second determination number; recording the number of the first target circular pattern or the second target circular pattern at the detection moment as a third determination number; Calculating image feature values ​​based on the first determination number, the second determination number, and the third determination number to obtain the image feature value at the detection moment; Wherein, after judging whether the target object has fallen based on the rotation angular velocity and the acceleration sensor data, the method further comprises: Based on the image feature value at each of the detection moments, a quantitative feature change value on a preset time series is obtained; wherein the preset time series is composed of a third number of the detection moments; When the quantity characteristic change value satisfies a fifth preset condition at a third target moment in the preset time sequence, and the quantity characteristic change value satisfies a sixth preset condition after the third target moment, it is determined that a fall event has occurred to the target object.

8. A fall protection vest, characterized in that: The fall protection vest comprises: A data acquisition module, used to acquire an environmental image and acceleration sensing data; wherein the environmental image and the acceleration sensing data correspond to the motion state of the target object at at least two detection moments; The central processing module includes a feature extraction software unit, a feature calculation software unit and a fall determination software unit, wherein the feature extraction software unit is used to extract line features for the environment image at each detection moment to obtain corresponding pattern line features; the feature calculation software unit is used to calculate the rotation angular velocity of the target object according to the pattern line features corresponding to at least two of the detection moments; wherein the pattern line features include ellipse feature data, straight line feature data, and circular feature data; two of the detection moments are divided into The method comprises: calculating the rotational angular velocity of the target object according to the pattern line features corresponding to at least two of the detection moments, comprising: obtaining first deflection data according to the ellipse feature data and the straight line feature data at the first target moment; obtaining second deflection data according to the ellipse feature data and the straight line feature data at the second target moment; calculating the rotational angular velocity of the target object based on the first deflection data, the second deflection data and the circular feature data; and the fall determination software unit is used to determine whether the target object falls based on the rotational angular velocity and the acceleration sensor data. The fall processing module is used to protect the target object when a fall event occurs to the target object.

9. The fall protection vest according to claim 8, characterized in that: The data acquisition module includes a first camera device and a second camera device, wherein the first camera device is arranged on the front side of the fall protection vest, and the second camera device is arranged on the shoulder side of the fall protection vest; wherein the fall protection vest includes block airbags; the block airbags include head and neck airbags, back airbags, left hip and hip airbags, and right hip and hip airbags.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a software program, and the processor implements the fall detection method according to any one of claims 1 to 7 when executing the software program.

Images