Method and system for identifying illegal behaviors of high-altitude operation

Abstract

The invention provides a high-altitude operation illegal behavior identification method and system, and relates to the technical field of monitoring image processing. The identification method comprises the following steps: acquiring target image information acquired by monitoring equipment; performing image processing on the human body recognition area to obtain an expanded human body recognition area with a preset pixel value; according to the human body key points, the extended human body identification area and the first safety device identification area, determining whether an illegal behavior of the first safety device exists; according to the maximum pixel height of the expanded human body recognition area and the second safety device contour recognition area, determining whether an illegal behavior of the second safety device exists or not; and carrying out image detection and identification on the identification area of the third safety device, and determining whether an illegal behavior of the third safety device exists or not according to a connection relationship between the third safety device and the target connection part. The high-altitude operation monitoring area can be covered by arranging a single monitoring device, and the method has the advantages of being low in cost and high in recognition precision.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring image processing, and in particular to a method and a system for identifying illegal behaviors in aerial work. Background Art

[0002] Violations in aerial work refer to the failure to comply with relevant safety regulations and operating procedures when performing aerial work (such as not wearing safety equipment, violating operating procedures, etc.). These violations may lead to serious safety accidents, so strict supervision is required. The identification method of aerial work violations is an important measure to monitor the safety of aerial work. It can detect violations in time and issue alarms.

[0003] However, the inventors found that the current methods for identifying violations of aerial work are generally based on the image recognition of safety devices by surveillance cameras, and the effective recognition distance of the surveillance cameras is short (for example, less than 5 meters). When the monitoring distance exceeds a certain distance, there is a problem of missed reports. In addition, in actual aerial work scenarios, some safety devices are easily blocked during the operation process, which may lead to false alarms. Summary of the invention

[0004] The present invention provides a method and system for identifying violations of aerial work, which are used to solve the problem of poor recognition accuracy in the current methods for identifying violations of aerial work.

[0005] One aspect of the present invention provides a method for identifying violations of aerial work, including: acquiring target image information collected by a monitoring device; performing image detection and identification on the target image information to determine a human body recognition area; performing image processing on the human body recognition area to obtain an extended human body recognition area with a preset pixel value; performing image detection and identification on the extended human body recognition area to at least determine human body key points, a first safety device recognition area, a second safety device contour recognition area, and a third safety device recognition area; determining whether there is a violation of the first safety device based on the human body key points, the extended human body recognition area, and the first safety device recognition area; determining whether there is a violation of the second safety device based on the maximum pixel height of the extended human body recognition area and the second safety device contour recognition area; performing image detection and identification on the third safety device recognition area, and determining whether there is a violation of the third safety device based on the connection relationship between the third safety device and the target connection part.

[0006] According to some embodiments of the present invention, performing image processing on a human body recognition area to obtain an extended human body recognition area with preset pixel values ​​includes: determining the pixel coordinates of the center point of a human body in the human body recognition area; performing image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the target image information after tracking; performing pixel scaling processing on the target image information after tracking so that the pixel value of the target image information after tracking is a preset pixel value; performing image magnification processing on the human body recognition area in the target image information with a preset pixel value based on a first preset frame expansion ratio to obtain an extended human body recognition area.

[0007] According to some embodiments of the present invention, determining whether there is a violation of the first safety device based on human body key points, an extended human body recognition area and a first safety device recognition area includes: determining a human waist key point based on the human body key points; determining a first human body area and a second human body area based on the human waist key point and the extended human body recognition area; determining a first pixel area ratio between the first human body area and the first safety device recognition area; determining a second pixel area ratio between the second human body area and the first safety device recognition area; determining that there is a violation of the first safety device in the target image information of the current frame when the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than a preset threshold; traversing multiple consecutive frames of image information of the target image information; and determining that there is a violation of the first safety device when the number of images in which the violation of the first safety device exists in the multiple consecutive frames of image information exceeds a preset number of frames.

[0008] According to some embodiments of the present invention, determining whether there is a violation of the second safety device based on the maximum pixel heights of the extended human body recognition area and the second safety device contour recognition area includes: determining the maximum pixel height of the human body based on the extended human body recognition area; determining the maximum pixel height of the second safety device based on the second safety device contour recognition area; determining that there is a violation of the second safety device in the target image information of the current frame when the maximum pixel height of the second safety device is less than the maximum pixel height of the human body; traversing multiple consecutive frames of image information of the target image information; and determining that there is a violation of the second safety device when the number of images in which the second safety device violates the law in the multiple consecutive frames of image information exceeds a preset number of frames.

[0009] According to some embodiments of the present invention, the target connection part includes at least a first target connection part and a second target connection part, and image detection and identification are performed on a third safety device identification area. Determining whether there is any violation of the third safety device based on the connection relationship between the third safety device and the target connection part includes: performing image enlargement processing on the third safety device identification area based on a second preset expansion frame ratio to obtain an expanded third safety device identification area; judging the connection relationship between the third safety device and the first target connection part and the second target connection part based on the expanded third safety device identification area; and determining that there is no violation of the third safety device when one end of the third safety device is connected to the first target connection part and the other end is connected to the second target connection part.

[0010] One aspect of the present invention provides a system for identifying violations of aerial work, including an image acquisition module, a region recognition module and a behavior determination module. The image acquisition module acquires target image information collected by a monitoring device; the region recognition module performs image detection and recognition on the target image information to determine a human body recognition area, performs image processing on the human body recognition area to obtain an extended human body recognition area with a preset pixel value, and performs image detection and recognition on the extended human body recognition area to at least determine human body key points, a first safety device recognition area, a second safety device contour recognition area and a third safety device recognition area; the behavior determination module determines whether there is a violation of the first safety device based on the human body key points, the extended human body recognition area and the first safety device recognition area, determines whether there is a violation of the second safety device based on the maximum pixel height of the extended human body recognition area and the second safety device contour recognition area, and performs image detection and recognition on the third safety device recognition area, and determines whether there is a violation of the third safety device based on the connection relationship between the third safety device and the target connection part.

[0011] According to some embodiments of the present invention, the region recognition module determines the pixel coordinates of the center point of the human body in the human body recognition area; the region recognition module performs image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the target image information after tracking; the region recognition module performs pixel scaling processing on the target image information after tracking so that the pixel value of the target image information after tracking is a preset pixel value; the region recognition module performs image enlargement processing on the human body recognition area in the target image information with a preset pixel value based on a first preset frame expansion ratio to obtain an expanded human body recognition area.

[0012] According to some embodiments of the present invention, the behavior recognition module determines the human waist key point based on the human body key point; the behavior recognition module determines the first human body area and the second human body area based on the human waist key point and the extended human body recognition area; the behavior recognition module determines the first pixel area ratio of the first human body area to the first safety device recognition area; the behavior recognition module determines the second pixel area ratio of the second human body area to the first safety device recognition area; the behavior recognition module determines that there is a violation of the first safety device in the target image information of the current frame when the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than a preset threshold; the behavior recognition module traverses consecutive multiple frames of image information of the target image information; the behavior recognition module determines that there is a violation of the first safety device when the number of images containing the violation of the first safety device in the consecutive multiple frames of image information exceeds a preset number of frames.

[0013] According to some embodiments of the present invention, the behavior recognition module determines the maximum pixel height of the human body based on the extended human body recognition area; the behavior recognition module determines the maximum pixel height of the second safety device based on the second safety device contour recognition area; the behavior recognition module determines that there is a violation of the second safety device in the target image information of the current frame when the maximum pixel height of the second safety device is less than the maximum pixel height of the human body; the behavior recognition module traverses consecutive multiple frames of image information of the target image information; the behavior recognition module determines that there is a violation of the second safety device when the number of images in which there is a violation of the second safety device in the consecutive multiple frames of image information exceeds a preset number of frames.

[0014] According to some embodiments of the present invention, the target connection part includes at least a first target connection part and a second target connection part, and the behavior recognition module performs image magnification processing on the third safety device recognition area based on a second preset expansion frame ratio to obtain an extended third safety device recognition area; the behavior recognition module determines the connection relationship between the third safety device and the first target connection part and the second target connection part based on the extended third safety device recognition area; the behavior recognition module determines that there is no violation of the third safety device when one end of the third safety device is connected to the first target connection part and the other end is connected to the second target connection part.

[0015] The present invention can obtain the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area respectively by performing image detection and identification on the extended human body identification area. And by performing image detection and identification on the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area, it is possible to accurately identify whether there is a violation of the first safety device, a violation of the second safety device and a violation of the first safety device. The present invention can cover the high-altitude operation monitoring area by setting a single monitoring device, and has the characteristics of low cost and high recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technical personnel in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flow chart showing a method for identifying violations of aerial work according to an embodiment of the present invention;

[0018] Figure 2 Another flow chart showing the method for identifying violations of aerial work according to an embodiment of the present invention;

[0019] Figure 3 A schematic diagram showing key points of a human body according to an embodiment of the present invention;

[0020] Figure 4 Another flow chart showing the method for identifying violations of aerial work according to an embodiment of the present invention;

[0021] Figure 5 Another flow chart showing the method for identifying violations of aerial work according to an embodiment of the present invention;

[0022] Figure 6 Another flow chart showing the method for identifying violations of aerial work according to an embodiment of the present invention;

[0023] Figure 7 A schematic diagram showing a system for identifying violations of aerial work according to an embodiment of the present invention.

[0024] Description of reference numerals:

[0025] A system for identifying illegal behaviors of aerial work 1; an image acquisition module 10; a region identification module 20; and a behavior identification module 30. DETAILED DESCRIPTION

[0026] The following is a clear and complete description of the technical solution of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by the skilled person in the art without creative work are within the scope of protection of the present invention.

[0027] The full names of the English abbreviations involved in the present invention and their Chinese meanings are:

[0028] SDK: Software Development Kit;

[0029] RTSP: Real Time Streaming Protocol, real-time streaming protocol;

[0030] Deeplab: image semantic segmentation model;

[0031] Mask2Former:image segmentation model;

[0032] SegFormer: semantic segmentation model.

[0033] OpenPose: Human pose estimation technology based on deep learning;

[0034] AlphaPose - a human pose estimation library based on deep learning technology.

[0035] HRNet: High-ResolutionNet, high-resolution deep neural network.

[0036] According to one aspect of the present invention, the present invention provides a method for identifying violations of high-altitude operations, which is used to identify whether there are violations of high-altitude operations in the current monitoring screen based on image information collected by a monitoring device.

[0037] Figure 1 A flow chart of a method for identifying violations of aerial work according to an embodiment of the present invention is shown. Figure 1 As shown, the method for identifying violations of aerial work may include steps S100 to S700. Exemplarily, the method for identifying violations of aerial work may be executed by a system for identifying violations of aerial work with computing capabilities.

[0038] According to an example embodiment, in step S100, the system for identifying violations of aerial work obtains image information collected by monitoring equipment.

[0039] For example, the monitoring device may be a high-altitude monitoring ball camera. The high-altitude monitoring ball camera includes an integrated camera and a pan-tilt mechanism. Based on the integrated camera and the pan-tilt mechanism, the high-altitude monitoring ball camera can achieve 360° horizontal rotation and 90° vertical tilt, can cover a wide monitoring area, and realize all-round monitoring without blind spots.

[0040] Exemplarily, the high-altitude monitoring ball camera can be installed at a preset position according to user needs. Exemplarily, the preset position can be a position with a certain monitoring height (such as a high position on a power tower, etc.).

[0041] The resolution of the high-altitude surveillance ball camera can be above 1080p, and multiple cruise points can be set through the pan-tilt mechanism according to user needs. After a preset interval, the high-altitude surveillance ball camera can automatically switch to other cruise points. At different cruise points, the high-altitude surveillance ball camera can set different resolutions and focal lengths.

[0042] In addition, the high-altitude monitoring ball camera can keep the integrated camera at a horizontal or upward angle at the cruising point. Such a setting can make the monitoring screen of the high-altitude monitoring ball camera cover the screen in the air instead of covering the screen on the ground.

[0043] The aerial work violation identification system can communicate with the aerial surveillance camera, and by calling the SDK of the aerial surveillance camera and the preset video stream function interface, it can obtain the RTSP video stream from the aerial surveillance camera. The aerial work violation identification system performs stream decoding processing on the RTSP video to obtain target image information including the target monitoring area (such as the aerial work area).

[0044] Exemplarily, the target image information may be color image data or grayscale image data.

[0045] In step S200, the aerial work violation identification system performs image detection and identification on the target image information to determine the human body identification area.

[0046] For example, the aerial work violation identification system can identify the coordinates of a series of points in the target image information, and generate a polygonal area based on the coordinates of these points to form an identification area.

[0047] The aerial work violation identification system can perform human body detection on the target image information based on human body segmentation technology. For example, the aerial work violation identification system can obtain the pixel coordinates of the rectangular frame of the human body area, and the human body identification area can be obtained according to the pixel coordinates of the rectangular frame.

[0048] Exemplarily, human body segmentation technology includes but is not limited to Deeplab, Mask2Former or SegFormer, etc., and the present invention is not limited to this.

[0049] In step S300, the aerial work violation identification system performs image processing on the human body identification area to obtain an extended human body identification area with a preset pixel value.

[0050] For example, image processing includes but is not limited to human body center tracking processing, pixel value processing based on image scaling, and magnification processing.

[0051] After the aerial work violation identification system performs image processing on the human body identification area, a new human body identification area with a preset pixel value can be obtained after image tracking and amplification processing of the human body identification area.

[0052] It can be understood that after the aerial work violation identification system completes the human body identification and tracking at the current cruise point based on the aerial surveillance camera, it can automatically switch to the next preset cruise point and perform human body identification and tracking at the next preset cruise point again. With this setting, the aerial work violation identification system can identify all human body identification areas in the monitoring area.

[0053] Figure 2 Another flow chart of the method for identifying violations of aerial work according to an embodiment of the present invention is shown.

[0054] Alternatively, if Figure 2 As shown, step S300 may also include steps S310-S340.

[0055] In step S310, the aerial work violation identification system determines the pixel coordinates of the center point of the human body in the human body identification area.

[0056] In step S320, the aerial work violation identification system performs image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the tracked target image information.

[0057] For example, the aerial work violation identification system can determine the pixel coordinates of the center of the human body based on the pixel coordinates of the rectangular frame of the human body area. The aerial work violation identification system can control the rotation of the aerial monitoring ball camera by calling the SDK of the aerial monitoring ball camera to adjust the input target monitoring area, so that the pixel coordinates of the center of the human body can always be set at the center of the monitoring screen.

[0058] In step S330, the aerial work violation identification system performs pixel scaling processing on the tracked target image information so that the pixel value of the tracked target image information is a preset pixel value.

[0059] For example, the high-altitude work violation identification system can use the high-altitude monitoring ball camera to zoom in and out the target image information to adjust the pixel size of the human body recognition area.

[0060] Exemplarily, when the pixel value of the human body recognition area is greater than the preset pixel value, the high-altitude work violation identification system can use the high-altitude monitoring ball camera to reduce the target image information so that the pixel value of the human body recognition area is reduced to the preset pixel value; or, when the pixel value of the human body recognition area is less than the preset pixel value, the high-altitude work violation identification system can use the high-altitude monitoring ball camera to enlarge the target image information so that the pixel width and height of the human body recognition area are enlarged to the preset pixel width and height.

[0061] With such a configuration, the present invention can adjust the pixel size of the human body recognition area so that the pixel width and height of the human body recognition area reach the most accurate preset pixel width and height for recognition, thereby preventing false detection and missed detection of human bodies, thereby improving the recognition accuracy of the present invention.

[0062] In step S340, the aerial work violation identification system performs image magnification processing on the human body recognition area in the target image information with preset pixel values ​​based on a first preset frame expansion ratio to obtain an expanded human body recognition area.

[0063] For example, the high-altitude work violation identification system tracks the target image information based on the pixel coordinates of the center point of the human body, so that the human body position and the preset pixel value can always be kept centered within the preset duration in the target image information, thereby obtaining a new human body identification area.

[0064] The aerial work violation identification system performs frame expansion processing on the new human body recognition area. That is, according to the pixel coordinates of the rectangular frame of the new human body recognition area, the frame is expanded in the horizontal direction and the vertical direction based on the first preset frame expansion ratio, and the expanded human body recognition area can be obtained according to the pixel coordinates after the frame expansion.

[0065] Exemplarily, the first preset frame expansion ratio can be customized according to user needs.

[0066] In this way, the present invention can make the first safety device, the second safety device and the third safety device all identifiable in the expanded frame area through the frame expansion and magnification processing, and make the first safety device, the second safety device and the third safety device identifiable in most areas in the expanded frame area as much as possible. The problem of false positives or false negatives caused by the safety device being blocked can be avoided, thereby improving the recognition accuracy of the present invention.

[0067] Figure 3 A schematic diagram showing key points of a human body according to an embodiment of the present invention.

[0068] In step S400, the aerial work violation identification system performs image detection and identification on the extended human body identification area to at least determine the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area.

[0069] For example, the first safety device identification area may be a seat belt identification area, the second safety device contour identification area may be a safety rope identification area, and the third safety device identification area may be an auto-locking device identification area or a differential identification area.

[0070] The high-altitude work violation identification system can expand the human body recognition area based on the human body key point detection technology to obtain the human body key points (such as Figure 3 ). Human key point detection technology includes but is not limited to OpenPose, AlphaPose or HRNet, etc., and the present invention is not limited to this.

[0071] The system for identifying violations of aerial work can obtain the pixel coordinates of a series of points on the outline of the safety rope based on the extended human body recognition area, thereby determining the safety rope recognition area; and the system for identifying violations of aerial work can also determine the pixel coordinates of the rectangular frame of the self-locking device or the differential based on the extended human body recognition area, thereby determining the self-locking device recognition area or the differential recognition area.

[0072] In step S500, the aerial work violation identification system determines whether there is any violation of the first safety device according to the key points of the human body, the extended human body identification area and the first safety device identification area.

[0073] For example, the violation of the first safety device may be failure to wear a safety belt. The aerial work violation identification system can determine the key feature points of the waist of the human body based on the key points of the human body, thereby determining the segmented area of ​​the human body based on the key feature points of the waist of the human body and the extended human body identification area, and determining whether the person is wearing a safety belt based on the image association between the segmented area of ​​the human body and the identification area of ​​the first safety device.

[0074] Figure 4 Another flow chart of the method for identifying violations of aerial work according to an embodiment of the present invention is shown.

[0075] Alternatively, if Figure 4 As shown, step S500 may also include steps S510-S570.

[0076] In step S510, the aerial work violation identification system determines the key points of the human waist based on the key points of the human body.

[0077] For example, the high-altitude work violation identification system can detect the waist key points (such as Figure 3 The human body key points 8 and human body key 11 shown in FIG.

[0078] In step S520, the aerial work violation identification system determines the first human body area and the second human body area based on the key points of the human waist and the extended human body identification area.

[0079] For example, the first human body region may be the upper body region of the human body, and the second human body region may be the lower body region of the human body. The aerial work violation identification system may use the human waistline (i.e., the line between the human body key point 8 and the human body key point 11) as the dividing line, and may identify the upper body region and the lower body region of the human body in combination with the expanded human body identification region.

[0080] In step S530, the aerial work violation identification system determines a first pixel area ratio between the first human body area and the first safety device identification area.

[0081] In step S540, the aerial work violation identification system determines a second pixel area ratio between the second human body area and the first safety device identification area.

[0082] In step S550, the aerial work violation identification system determines that there is a violation of the first safety device in the target image information of the current frame when the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than a preset threshold.

[0083] For example, the pixel area ratio refers to the ratio of the number of pixels in a certain area of ​​the image to the total number of pixels in the entire image. By calculating the ratio of the number of pixels in different areas of the image, the area ratio of these areas in the image can be obtained.

[0084] The high-altitude work violation identification system calculates the first pixel area ratio between the upper body area and the safety belt identification area to determine whether the safety belt is worn in the upper body area of ​​the human body; and the high-altitude work violation identification system calculates the second pixel area ratio between the lower body area and the safety belt identification area to determine whether the safety belt is located in the lower body area of ​​the human body.

[0085] When it is determined that the safety belt is not worn in the upper body area of ​​the human body, or when it is determined that the safety belt is located in the lower body area of ​​the human body, the high-altitude work violation identification system determines that there is a violation of not wearing a safety belt.

[0086] In step S560, the aerial work violation identification system traverses multiple consecutive frames of target image information.

[0087] In step S570, the aerial work violation identification system determines that there is a violation of the first safety device when the number of images of violations of the first safety device in the continuous multiple frames of image information exceeds a preset number of frames.

[0088] For example, when the high-altitude work violation identification system finds that there is a violation of not wearing a safety belt in the target image information of the current frame, it traverses the continuous N frames of the target image information. 1 Frame image. In the presence of M 1 The recognition results of the frame images are all violations of the first safety device, and M 1 / N 1 When the value is greater than a first preset threshold, the aerial work violation identification system determines that there is a violation of the first safety device and issues a corresponding alarm.

[0089] The present invention can perform dual detection and identification of the position of the seat belt by judging whether the seat belt is worn in the upper body area of ​​the human body and judging whether the seat belt is located in the lower body area of ​​the human body, and can accurately judge whether there is a violation of not wearing the seat belt.

[0090] In step S600, the aerial work violation identification system determines whether there is a violation of the second safety device according to the maximum pixel height of the extended human body identification area and the second safety device contour identification area.

[0091] For example, the maximum pixel height may be the pixel value in the vertical direction of the highest point of the target object. The violation of the second safety device may be not wearing a safety rope. The aerial work violation identification system may determine whether there is a violation of the safety rope based on the maximum pixel height of the safety rope and the maximum pixel height of the human body.

[0092] Figure 5 Another flow chart of the method for identifying violations of aerial work according to an embodiment of the present invention is shown.

[0093] Alternatively, if Figure 5 As shown, step S600 may also include steps S610-S650.

[0094] In step S610, the aerial work violation identification system determines the maximum pixel height of the human body according to the expanded human body identification area.

[0095] In step S620, the aerial work violation identification system determines the maximum pixel height of the second safety device according to the second safety device contour identification area.

[0096] In step S630, the aerial work violation identification system determines that there is a violation of the second safety device in the target image information of the current frame when the maximum pixel height of the second safety device is less than the maximum pixel height of the human body.

[0097] In step S640, the aerial work violation identification system traverses multiple consecutive frames of target image information.

[0098] In step S650, the aerial work violation identification system determines that there is a violation of the second safety device when the number of images showing a violation of the second safety device in a plurality of consecutive frames of image information exceeds a preset number of frames.

[0099] For example, the aerial work violation identification system can determine the rectangular frame of the safety rope outline and calculate the maximum pixel height of the safety rope; and the aerial work violation identification system can determine the maximum pixel height of the human body in the human body identification area. For example, the maximum pixel height can be the vertical pixel value of the upper left point or the upper right point of the target rectangular frame.

[0100] When the high-altitude work violation identification system determines that the safety rope is not worn and the maximum pixel height is less than the maximum pixel height of the human body, it determines that there is a violation of not wearing the safety rope.

[0101] For example, when the high-altitude work violation identification system finds that there is a violation of not wearing a safety rope in the target image information of the current frame, it traverses the continuous N frames of the target image information. 2 Frame image. In the presence of M 2 The recognition results of the frame images are all violations of the second safety device, and M 2 / N 2 When the value is greater than the second preset threshold, the aerial work violation identification system determines that there is a violation of the second safety device and issues a corresponding alarm.

[0102] The present invention can accurately determine the actual height of the safety rope and the human body by determining the maximum pixel height between the safety rope and the human body, and thus can accurately determine whether there is a violation of not wearing the safety rope based on the actual height between the safety rope and the human body.

[0103] In step S700, the aerial work violation identification system performs image detection and identification on the third safety device identification area, and determines whether there is any violation of the third safety device based on the connection relationship between the third safety device and the target connection part.

[0104] For example, the violation behavior of the third safety device may be an improper operation of the self-locking device or an improper operation of the differential.

[0105] The aerial work violation identification system can detect the connection relationship between the third safety device and other components based on image detection and recognition of the third safety device identification area. By identifying and detecting the connection relationship, it can be identified whether there is improper operation of the self-locking device or abnormal operation of the differential.

[0106] Figure 6 Another flow chart of the method for identifying violations of aerial work according to an embodiment of the present invention is shown.

[0107] Alternatively, if Figure 6 As shown, step S700 may further include steps S710-S730. The target connection part may be a component that is mechanically connected to the third safety device. The target connection part includes at least a first target connection part and a second target connection part.

[0108] In step S710, the aerial work violation identification system performs image magnification processing on the third safety device identification area based on a second preset frame expansion ratio to obtain an expanded third safety device identification area.

[0109] For example, the aerial work violation identification system can expand the self-locking device identification area or the differential identification area. That is, according to the pixel coordinates of the rectangular frame of the self-locking device identification area or the differential identification area, the frame is expanded in the horizontal direction and the vertical direction based on the second preset frame expansion ratio, and according to the pixel coordinates after the frame expansion, the rectangular frame image of the self-locking device or the differential after the frame expansion can be obtained (that is, the third safety device identification area is expanded).

[0110] Exemplarily, the second preset frame expansion ratio can be customized according to user needs.

[0111] In step S720, the aerial work violation identification system determines the connection relationship between the third safety device and the first target connection part and the second target connection part according to the expanded third safety device identification area.

[0112] In step S730, the aerial work violation identification system determines that there is no violation of the third safety device when one end of the third safety device is connected to the first target connection part and the other end is connected to the second target connection part.

[0113] For example, the aerial work violation identification system performs image recognition classification on the connection relationship between the third safety device and the target connection part based on image recognition, and the image recognition classification results can be qualified or unqualified.

[0114] Exemplarily, when the third safety device is a self-locking device, the first target connection part may be a safety belt, and the second target connection part may be a safety rope; when the third safety device is a differential, the first target connection part may be a safety rope, and the second target connection part may be a bracket (such as a tower bracket, etc.), and the present invention is not limited to this.

[0115] For example, when the third safety device is a self-locking device, the high-altitude work violation identification system uses an image classification algorithm (such as a preset training image classification model) to recognize that one end of the self-locking device is connected to the safety belt and the other end is connected to the safety rope. The image recognition classification of the self-locking device is qualified, otherwise it is unqualified (that is, there is an improper operation of the self-locking device).

[0116] When the high-altitude operation violation identification system fails to correctly operate the self-locking device in the target image information of the current frame, it traverses the continuous N 3 Frame image. In the presence of M 3 The recognition results of the frame images all show that the self-locking device is not operated correctly, and M 3 / N 3 When the value is greater than the third preset threshold, the aerial work violation identification system determines that there is a violation of the third safety device and issues a corresponding alarm.

[0117] For another example, when the third safety device is a differential, the aerial work violation identification system uses an image classification algorithm (such as a preset training image classification model) to identify that one end of the differential is connected to a safety rope and the other end is connected to an electric tower support. The image recognition classification of the differential is qualified, otherwise it is unqualified (i.e., there is improper operation of the differential).

[0118] When the high-altitude operation violation identification system does not correctly operate the differential behavior in the target image information of the current frame, it traverses the continuous N 4 Frame image. In the presence of M 4 The recognition results of the frame images all show that the differential is not operated correctly, and M 4 / N 4 When the value is greater than the fourth preset threshold, the aerial work violation identification system determines that there is a violation of the third safety device and issues a corresponding alarm.

[0119] The present invention is based on image detection and recognition technology. By judging the connection relationship between the third safety device and the target connection part, it can accurately judge whether the third safety device is operating normally, and thus can judge whether there is any violation of not operating the third safety device normally.

[0120] Through the above-mentioned example embodiments, the present invention can obtain the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area respectively by performing image detection and identification on the extended human body identification area. And through image detection and identification of the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area, it is possible to accurately identify whether there is a violation of the first safety device, a violation of the second safety device and a violation of the first safety device. The present invention can cover the high-altitude operation monitoring area by setting a single monitoring device, and has the characteristics of low cost and high recognition accuracy.

[0121] According to one aspect of the present invention, the present invention provides a system for identifying violations of aerial work, which is used to identify whether there are violations of aerial work in the current monitoring screen based on image information collected by a monitoring device.

[0122] Figure 7 A schematic diagram of a system for identifying violations of aerial work according to an embodiment of the present invention is shown. Figure 7 As shown, the aerial work violation behavior identification system 1 includes an image acquisition module 10 , a region identification module 20 and a behavior identification module 30 .

[0123] According to an example embodiment, the image acquisition module 10 acquires image information collected by a monitoring device.

[0124] For example, the monitoring device may be a high-altitude monitoring ball camera. The high-altitude monitoring ball camera includes an integrated camera and a pan-tilt mechanism. Based on the integrated camera and the pan-tilt mechanism, the high-altitude monitoring ball camera can achieve 360° horizontal rotation and 90° vertical tilt, can cover a wide monitoring area, and realize all-round monitoring without blind spots.

[0125] Exemplarily, the high-altitude monitoring ball camera can be installed at a preset position according to user needs. Exemplarily, the preset position can be a position with a certain monitoring height (such as a high position on a power tower, etc.).

[0126] The resolution of the high-altitude surveillance ball camera can be above 1080p, and multiple cruise points can be set through the pan-tilt mechanism according to user needs. After a preset interval, the high-altitude surveillance ball camera can automatically switch to other cruise points. At different cruise points, the high-altitude surveillance ball camera can set different resolutions and focal lengths.

[0127] In addition, the high-altitude monitoring ball camera can keep the integrated camera at a horizontal or upward angle at the cruising point. Such a setting can make the monitoring screen of the high-altitude monitoring ball camera cover the screen in the air instead of covering the screen on the ground.

[0128] The image acquisition module 10 can be connected to the high-altitude monitoring ball camera for communication, and can obtain the RTSP video stream from the high-altitude monitoring ball camera by calling the SDK of the high-altitude monitoring ball camera and the preset video stream function interface. The image acquisition module 10 performs stream decoding processing on the RTSP video to obtain target image information including the target monitoring area (such as the high-altitude working area).

[0129] Exemplarily, the target image information may be color image data or grayscale image data.

[0130] The area recognition module 20 performs image detection and recognition on the target image information to determine a human body recognition area.

[0131] For example, the region recognition module 20 may recognize the coordinates of a series of points in the target image information, and generate a polygonal region according to the coordinates of these points to form a recognition region.

[0132] The region recognition module 20 can perform human body detection on the target image information based on human body segmentation technology. For example, the region recognition module 20 can obtain the pixel coordinates of the rectangular frame of the human body region, and obtain the human body recognition region according to the pixel coordinates of the rectangular frame.

[0133] Exemplarily, human body segmentation technology includes but is not limited to Deeplab, Mask2Former or SegFormer, etc., and the present invention is not limited to this.

[0134] The region recognition module 20 performs image processing on the human body recognition region to obtain an extended human body recognition region with a preset pixel value.

[0135] For example, image processing includes but is not limited to human body center tracking processing, pixel value processing based on image scaling, and magnification processing.

[0136] After the region recognition module 20 performs image processing on the human body recognition region, a new human body recognition region with a preset pixel value can be obtained after image tracking and amplification processing of the human body recognition region.

[0137] It can be understood that after the area recognition module 20 completes the human body recognition tracking at the current cruise point based on the high-altitude surveillance camera, it can automatically switch to the next preset cruise point and perform human body recognition tracking at the next preset cruise point again. In this way, the area recognition module 20 can identify all human body recognition areas in the monitoring area.

[0138] Optionally, the region identification module 20 determines the pixel coordinates of the center point of the human body in the human body identification region.

[0139] The region identification module 20 performs image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the tracked target image information.

[0140] For example, the region identification module 20 can determine the pixel coordinates of the center point of the human body according to the pixel coordinates of the rectangular frame of the human body region. The region identification module 20 can control the rotation of the high-altitude monitoring ball camera by calling the SDK of the high-altitude monitoring ball camera to adjust the input target monitoring area, so that the pixel coordinates of the center point of the human body can always be set at the center position of the monitoring screen.

[0141] The region identification module 20 performs pixel scaling processing on the tracked target image information so that the pixel value of the tracked target image information is a preset pixel value.

[0142] For example, the area recognition module 20 can enlarge and reduce the target image information through a high-altitude monitoring ball camera to adjust the pixel size of the human body recognition area.

[0143] Exemplarily, when the pixel value of the human body recognition area is greater than the preset pixel value, the area recognition module 20 can use the high-altitude monitoring ball camera to reduce the target image information so that the pixel value of the human body recognition area is reduced to the preset pixel value; or, when the pixel value of the human body recognition area is less than the preset pixel value, the area recognition module 20 can use the high-altitude monitoring ball camera to enlarge the target image information so that the pixel width and height of the human body recognition area are enlarged to the preset pixel width and height.

[0144] With such a configuration, the present invention can adjust the pixel size of the human body recognition area so that the pixel width and height of the human body recognition area reach the most accurate preset pixel width and height for recognition, thereby preventing false detection and missed detection of human bodies, thereby improving the recognition accuracy of the present invention.

[0145] The region recognition module 20 performs image magnification processing on the human body recognition region in the target image information having a preset pixel value based on a first preset frame expansion ratio to obtain an expanded human body recognition region.

[0146] For example, the region identification module 20 tracks the target image information based on the pixel coordinates of the center point of the human body, so that the human body position and the preset pixel value can always be kept centered within a preset duration in the target image information, thereby obtaining a new human body identification region.

[0147] The region recognition module 20 performs frame expansion processing on the new human body recognition region, that is, according to the pixel coordinates of the rectangular frame of the new human body recognition region, the frame is expanded in the horizontal direction and the vertical direction based on the first preset frame expansion ratio, and the expanded human body recognition region can be obtained according to the pixel coordinates after the frame expansion.

[0148] Exemplarily, the first preset frame expansion ratio can be customized according to user needs.

[0149] In this way, the present invention can make the first safety device, the second safety device and the third safety device all identifiable in the expanded frame area through the frame expansion and magnification processing, and make the first safety device, the second safety device and the third safety device identifiable in most areas in the expanded frame area as much as possible. The problem of false positives or false negatives caused by the safety device being blocked can be avoided, thereby improving the recognition accuracy of the present invention.

[0150] According to an exemplary embodiment, the area recognition module 20 performs image detection and recognition on the extended human body recognition area to at least determine human body key points, a first safety device recognition area, a second safety device contour recognition area, and a third safety device recognition area.

[0151] For example, the first safety device identification area may be a seat belt identification area, the second safety device contour identification area may be a safety rope identification area, and the third safety device identification area may be an auto-locking device identification area or a differential identification area.

[0152] The region recognition module 20 can obtain the key points of the human body (such as Figure 3 ). Human key point detection technology includes but is not limited to OpenPose, AlphaPose or HRNet, etc., and the present invention is not limited to this.

[0153] The area identification module 20 can obtain the pixel coordinates of a series of points on the safety rope contour based on the extended human body identification area, thereby determining the safety rope identification area; and the area identification module 20 can also determine the pixel coordinates of the rectangular frame of the self-locking device or the differential based on the extended human body identification area, thereby determining the self-locking device identification area or the differential identification area.

[0154] The behavior recognition module 30 determines whether there is any violation of the first safety device according to the key points of the human body, the extended human body recognition area and the first safety device recognition area.

[0155] For example, the violation behavior of the first safety device may be failure to wear a seat belt. The behavior recognition module 30 may determine the waist key feature points of the human body according to the key points of the human body, thereby determining the segmented area of ​​the human body according to the waist key feature points of the human body and the extended human body recognition area, and determining whether the person is wearing a seat belt according to the image association between the segmented area of ​​the human body and the recognition area of ​​the first safety device.

[0156] Optionally, the behavior recognition module 30 determines the key points of the waist of the human body according to the key points of the human body.

[0157] For example, the behavior recognition module 30 can detect waist key points (such as Figure 3 The human body key points 8 and human body key 11 shown in FIG.

[0158] The behavior recognition module 30 determines the first human body region and the second human body region according to the key points of the human waist and the extended human body recognition region.

[0159] For example, the first human body region may be the upper body region of the human body, and the second human body region may be the lower body region of the human body. The behavior recognition module 30 may use the human waistline (i.e., the line between the human body key point 8 and the human body key point 11) as a dividing line, and in combination with the expanded human body recognition region, the upper body region and the lower body region of the human body may be recognized.

[0160] The behavior recognition module 30 determines a first pixel area ratio between the first human body region and the first safety device recognition region.

[0161] The behavior recognition module 30 determines a second pixel area ratio between the second human body region and the first safety device recognition region.

[0162] The behavior recognition module 30 determines that there is a violation of the first safety device in the target image information of the current frame when the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than the preset threshold.

[0163] For example, the pixel area ratio refers to the ratio of the number of pixels in a certain area of ​​the image to the total number of pixels in the entire image. By calculating the ratio of the number of pixels in different areas of the image, the area ratio of these areas in the image can be obtained.

[0164] The behavior recognition module 30 can determine whether the seat belt is worn in the upper body area of ​​the human body by calculating the first pixel area ratio between the upper body area of ​​the human body and the seat belt identification area; and the behavior recognition module 30 can determine whether the seat belt is located in the lower body area of ​​the human body by calculating the second pixel area ratio between the lower body area of ​​the human body and the seat belt identification area.

[0165] When it is determined that the seat belt is not worn in the upper body area of ​​the human body, or when it is determined that the seat belt is located in the lower body area of ​​the human body, the behavior recognition module 30 determines that there is a violation behavior of not wearing the seat belt.

[0166] The behavior recognition module 30 traverses multiple consecutive frames of target image information.

[0167] The behavior recognition module 30 determines that there is a violation of the first safety device when the number of images containing the violation of the first safety device in the continuous plurality of frames of image information exceeds a preset number of frames.

[0168] For example, when the target image information of the current frame contains a violation of not wearing a seat belt, the behavior recognition module 30 traverses the continuous N 1 Frame image. In the presence of M 1 The recognition results of the frame images are all violations of the first safety device, and M 1 / N 1 When the value is greater than the first preset threshold, the behavior recognition module 30 determines that there is a violation of the first safety device and issues a corresponding alarm.

[0169] The present invention can perform dual detection and identification of the position of the seat belt by judging whether the seat belt is worn in the upper body area of ​​the human body and judging whether the seat belt is located in the lower body area of ​​the human body, and can accurately judge whether there is a violation of not wearing the seat belt.

[0170] According to an example embodiment, the behavior recognition module 30 determines whether there is a violation of the second safety device according to the maximum pixel heights of the extended human body recognition area and the second safety device contour recognition area.

[0171] For example, the maximum pixel height may be the pixel value in the vertical direction of the highest point of the target object. The violation of the second safety device may be not wearing a safety rope. The behavior recognition module 30 may determine whether there is a violation of not wearing a safety rope based on the maximum pixel height of the safety rope and the maximum pixel height of the human body.

[0172] Optionally, the behavior recognition module 30 determines the maximum pixel height of the human body according to the expanded human body recognition area.

[0173] The behavior recognition module 30 determines the maximum pixel height of the second safety device according to the second safety device contour recognition area.

[0174] When the maximum pixel height of the second safety device is less than the maximum pixel height of the human body, the behavior recognition module 30 determines that there is a violation of the second safety device in the target image information of the current frame.

[0175] The behavior recognition module 30 traverses multiple consecutive frames of target image information.

[0176] The behavior recognition module 30 determines that there is a violation of the second safety device when the number of images containing the violation of the second safety device in the continuous multiple frames of image information exceeds a preset number of frames.

[0177] For example, the behavior recognition module 30 can determine the rectangular frame of the safety rope outline and calculate the maximum pixel height of the safety rope; and the behavior recognition module 30 can determine the maximum pixel height of the human body in the human body recognition area. Exemplarily, the maximum pixel height can be the vertical pixel value of the upper left point or the upper right point of the target rectangular frame.

[0178] When the behavior recognition module 30 determines that the maximum pixel height of the safety rope is less than the maximum pixel height of the human body, it determines that there is a violation of not wearing the safety rope.

[0179] For example, when the target image information of the current frame contains a violation of not wearing a safety rope, the behavior recognition module 30 traverses the continuous N 2 Frame image. In the presence of M 2 The recognition results of the frame images are all violations of the second safety device, and M 2 / N 2 When the value is greater than the second preset threshold, the behavior recognition module 30 determines that there is a violation of the second safety device and issues a corresponding alarm.

[0180] The present invention can accurately determine the actual height of the safety rope and the human body by determining the maximum pixel height between the safety rope and the human body, and thus can accurately determine whether there is a violation of not wearing the safety rope based on the actual height between the safety rope and the human body.

[0181] According to an exemplary embodiment, the behavior recognition module 30 performs image detection and recognition on the third safety device recognition area, and determines whether there is any violation of the third safety device according to the connection relationship between the third safety device and the target connection part.

[0182] For example, the violation behavior of the third safety device may be an improper operation of the self-locking device or an improper operation of the differential.

[0183] The behavior recognition module 30 can detect the connection relationship between the third safety device and other components based on image detection and recognition of the third safety device recognition area. By identifying and detecting the connection relationship, it can be identified whether there is improper operation of the self-locking device or abnormal operation of the differential.

[0184] Optionally, the target connection part may be a component that is mechanically connected to the third safety device. The target connection part includes at least a first target connection part and a second target connection part.

[0185] The behavior recognition module 30 performs image magnification processing on the third safety device recognition area based on the second preset frame expansion ratio to obtain an expanded third safety device recognition area.

[0186] For example, the behavior recognition module 30 may perform frame expansion processing on the self-locking device recognition area or the differential recognition area. That is, according to the pixel coordinates of the rectangular frame of the self-locking device recognition area or the differential recognition area, the frame is expanded in the horizontal direction and the vertical direction based on the second preset frame expansion ratio, and according to the pixel coordinates after the frame expansion, the rectangular frame image of the self-locking device or the differential after the frame expansion can be obtained (that is, the third safety device recognition area is expanded).

[0187] Exemplarily, the second preset frame expansion ratio can be customized according to user needs.

[0188] The behavior recognition module 30 determines the connection relationship between the third safety device and the first target connection part and the second target connection part according to the expanded third safety device recognition area.

[0189] The behavior recognition module 30 determines that there is no violation behavior of the third safety device when one end of the third safety device is connected to the first target connection part and the other end of the third safety device is connected to the second target connection part.

[0190] For example, the behavior recognition module 30 performs image recognition classification on the connection relationship between the third safety device and the target connection part based on image recognition, and the image recognition classification result may be qualified or unqualified.

[0191] Exemplarily, when the third safety device is a self-locking device, the first target connection part may be a safety belt, and the second target connection part may be a safety rope; when the third safety device is a differential, the first target connection part may be a safety rope, and the second target connection part may be a bracket (such as a tower bracket, etc.), and the present invention is not limited to this.

[0192] For example, in the case where the third safety device is a self-locking device, the behavior recognition module 30 uses an image classification algorithm (such as a preset training image classification model) to recognize that one end of the self-locking device is connected to the safety belt and the other end is connected to the safety rope. The image recognition classification of the self-locking device is qualified, otherwise it is unqualified (that is, there is a behavior of improper operation of the self-locking device).

[0193] When the behavior recognition module 30 does not correctly operate the self-locking device behavior in the target image information of the current frame, it traverses the continuous N 3 Frame image. In the presence of M 3 The recognition results of the frame images all show that the self-locking device is not operated correctly, and M 3 / N 3 When the value is greater than the third preset threshold, the behavior recognition module 30 determines that there is a violation of the third safety device and issues a corresponding alarm.

[0194] For another example, when the third safety device is a differential, the behavior recognition module 30 uses an image classification algorithm (such as a preset training image classification model) to recognize that one end of the differential is connected to a safety rope and the other end is connected to an electric tower support. The image recognition classification of the differential is qualified, otherwise it is unqualified (i.e., there is a behavior of improper operation of the differential).

[0195] When the differential behavior is not correctly operated in the target image information of the current frame, the behavior recognition module 30 traverses the continuous N 4 Frame image. In the presence of M 4 The recognition results of the frame images all show that the differential is not operated correctly, and M 4 / N 4 When the value is greater than the fourth preset threshold, the behavior recognition module 30 determines that there is a violation of the third safety device and issues a corresponding alarm.

[0196] The present invention is based on image detection and recognition technology. By judging the connection relationship between the third safety device and the target connection part, it can accurately judge whether the third safety device is operating normally, and thus can judge whether there is any violation of not operating the third safety device normally.

[0197] Through the above-mentioned example embodiments, the present invention can obtain the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area respectively by performing image detection and identification on the extended human body identification area. And through image detection and identification of the key points of the human body, the first safety device identification area, the second safety device contour identification area and the third safety device identification area, it is possible to accurately identify whether there is a violation of the first safety device, a violation of the second safety device and a violation of the first safety device. The present invention can cover the high-altitude operation monitoring area by setting a single monitoring device, and has the characteristics of low cost and high recognition accuracy.

[0198] According to another aspect of the present invention, an electronic device is provided, which includes: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors can implement the method for identifying violations of aerial work as described above.

[0199] According to another aspect of the present invention, a non-volatile computer-readable storage medium is provided, wherein a computer program is stored on the storage medium, and when the computer program is executed by a processor, the method for identifying violations of aerial work as described above can be implemented.

[0200] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions of the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for identifying violations of aerial work, characterized in that: include: Obtain target image information collected by monitoring equipment; Performing image detection and recognition on the target image information to determine a human body recognition area; Performing image processing on the human body recognition area to obtain an extended human body recognition area with a preset pixel value; Performing image detection and recognition on the extended human body recognition area to at least determine key points of the human body, a first safety device recognition area, a second safety device contour recognition area, and a third safety device recognition area; Determining whether there is a violation of the first safety device according to the key points of the human body, the extended human body recognition area, and the first safety device recognition area; determining whether there is a violation of the second safety device according to the maximum pixel heights of the extended human body recognition area and the second safety device contour recognition area; Image detection and recognition are performed on the third safety device recognition area, and whether there is any violation of the third safety device is determined based on the connection relationship between the third safety device and the target connection part.

2. The method for identifying violations of aerial work according to claim 1, characterized in that: The performing image processing on the human body recognition area to obtain an extended human body recognition area with a preset pixel value includes: Determine the pixel coordinates of the center point of the human body in the human body recognition area; Performing image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the tracked target image information; Performing pixel scaling processing on the tracked target image information so that the pixel value of the tracked target image information is the preset pixel value; The human body recognition area in the target image information having the preset pixel value is subjected to image enlargement processing based on a first preset frame expansion ratio to obtain the extended human body recognition area.

3. The method for identifying violations of aerial work according to claim 1, characterized in that: The determining whether there is a violation of the first safety device according to the human body key points, the extended human body recognition area and the first safety device recognition area comprises: Determine the key points of the waist of the human body according to the key points of the human body; Determine a first human body region and a second human body region according to the human body waist key point and the extended human body recognition region; Determining a first pixel area ratio between the first human body region and the first safety device recognition region; Determine a second pixel area ratio of the second human body area to the first safety device recognition area; When the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than a preset threshold, determining that there is a violation of the first safety device in the target image information of the current frame; Traversing continuous multiple frames of image information of the target image information; If the number of images containing the violation of the first safety device in the continuous plurality of frames of image information exceeds a preset number of frames, it is determined that the violation of the first safety device exists.

4. The method for identifying violations of aerial work according to claim 1, characterized in that: The determining whether there is a violation of the second safety device according to the maximum pixel height of the extended human body recognition area and the second safety device contour recognition area comprises: Determine the maximum pixel height of the human body according to the extended human body recognition area; determining a maximum pixel height of the second safety device according to the second safety device contour recognition area; When the maximum pixel height of the second safety device is less than the maximum pixel height of the human body, determining that there is a violation of the second safety device in the target image information of the current frame; Traversing continuous multiple frames of image information of the target image information; If the number of images containing the violation of the second safety device in the continuous plurality of frames of image information exceeds a preset number of frames, it is determined that the violation of the second safety device exists.

5. The method for identifying violations of aerial work according to claim 1, characterized in that: The target connection portion includes at least a first target connection portion and a second target connection portion, and performing image detection and recognition on the third safety device recognition area and determining whether there is a violation of the third safety device according to the connection relationship between the third safety device and the target connection portion includes: Performing image magnification processing on the third safety device identification area based on a second preset frame expansion ratio to obtain an expanded third safety device identification area; determining, according to the expanded third safety device identification area, a connection relationship between the third safety device and the first target connection portion and the second target connection portion; In a case where one end of the third safety device is connected to the first target connection portion and the other end is connected to the second target connection portion, it is determined that there is no violation of the third safety device.

6. A system for identifying violations of aerial work, characterized in that: include: An image acquisition module acquires target image information collected by the monitoring device; a region recognition module, performing image detection and recognition on the target image information to determine a human body recognition region, performing image processing on the human body recognition region to obtain an extended human body recognition region with a preset pixel value, and performing image detection and recognition on the extended human body recognition region to at least determine a human body key point, a first safety device recognition region, a second safety device contour recognition region, and a third safety device recognition region; A behavior determination module determines whether there is any violation of the first safety device based on the human key points, the extended human recognition area and the first safety device recognition area, determines whether there is any violation of the second safety device based on the maximum pixel height of the extended human recognition area and the second safety device contour recognition area, and performs image detection and recognition on the third safety device recognition area, and determines whether there is any violation of the third safety device based on the connection relationship between the third safety device and the target connection part.

7. The system for identifying violations of aerial work according to claim 6 is characterized in that: The area recognition module determines the pixel coordinates of the center point of the human body in the human body recognition area; The region recognition module performs image tracking on the pixel coordinates of the center point of the human body so that the pixel coordinates of the center point of the human body are located at the center of the tracked target image information; The region identification module performs pixel scaling processing on the tracked target image information so that the pixel value of the tracked target image information is the preset pixel value; The region recognition module performs image magnification processing on the human body recognition region in the target image information having the preset pixel value based on a first preset frame expansion ratio to obtain the extended human body recognition region.

8. The system for identifying violations of aerial work according to claim 6 is characterized in that: The behavior recognition module determines the key points of the waist of the human body according to the key points of the human body; The behavior recognition module determines a first human body region and a second human body region according to the human waist key point and the extended human body recognition region; The behavior recognition module determines a first pixel area ratio between the first human body area and the first safety device recognition area; The behavior recognition module determines a second pixel area ratio between the second human body area and the first safety device recognition area; The behavior recognition module determines that there is a violation of the first safety device in the target image information of the current frame when the first pixel area ratio is less than a preset threshold or the second pixel area ratio is greater than a preset threshold; The behavior recognition module traverses multiple consecutive frames of image information of the target image information; The behavior recognition module determines that there is a violation of the first safety device when the number of images containing the violation of the first safety device in the continuous multiple frames of image information exceeds a preset number of frames.

9. The system for identifying violations of aerial work according to claim 6, characterized in that: The behavior recognition module determines the maximum pixel height of the human body according to the extended human body recognition area; The behavior recognition module determines the maximum pixel height of the second safety device according to the second safety device contour recognition area; The behavior recognition module determines that there is a violation of the second safety device in the target image information of the current frame when the maximum pixel height of the second safety device is less than the maximum pixel height of the human body; The behavior recognition module traverses multiple consecutive frames of image information of the target image information; The behavior recognition module determines that there is a violation of the second safety device when the number of images containing the violation of the second safety device in the continuous multiple frames of image information exceeds a preset number of frames.

10. The system for identifying violations of aerial work according to claim 6, characterized in that: The target connection portion at least includes a first target connection portion and a second target connection portion, and the behavior recognition module performs image magnification processing on the third safety device recognition area based on a second preset frame expansion ratio to obtain an expanded third safety device recognition area; The behavior recognition module determines the connection relationship between the third safety device and the first target connection part and the second target connection part according to the extended third safety device recognition area; The behavior recognition module determines that there is no violation behavior of the third safety device when one end of the third safety device is connected to the first target connection part and the other end of the third safety device is connected to the second target connection part.

Images