A dust integrated intelligent management method, device and medium

Abstract

The application discloses a kind of dust integrated intelligent management method, equipment and medium, belong to artificial intelligence technical field, for solving existing stockyard dust management method exists hysteresis, it is difficult to carry out accurate tracking type spray management according to the real-time generation of smoke plume dust of operation vehicle, caused the technical problem of poor management effect and easy to waste a large amount of water resources.Method includes: the target image obtained by pan-tilt camera is carried out edge pixel extraction, and the fitting circumferential trajectory of target image is obtained;Combining the fitting circumferential trajectory with the rotation angle of pan-tilt camera, the visual detection range of pan-tilt camera is obtained;Through visual detection range, target pixel difference detection is carried out on smoke plume dust and stockyard vehicle in stockyard, and the three-dimensional coordinate position of smoke plume dust and stockyard vehicle is obtained;According to three-dimensional coordinate position, spray control is carried out on fog gun installed directly above pan-tilt camera, to spray dust fall in dust fall area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, and in particular to a dust integrated intelligent management method, device and medium. BACKGROUND

[0002] The existing stockyard dust management is that when the concentration monitor detects that the dust concentration in the shed reaches the preset upper threshold, the fog gun is started to spray dust suppression. Its management has obvious hysteresis. At the same time, when the fog gun is started, its spraying mode is horizontal reciprocating rotation according to the preset program, which cannot accurately track the dust target for spraying, and the management effect is often poor.

[0003] When the transport vehicle and the excavator are working, it is easy to cause a large amount of plume dust in the stockyard, which causes a lot of pollution to the environment, and it is difficult to accurately position the vehicle during driving or operation, causing hysteresis of the plume dust management, making the dust management efficiency low, and wasting a lot of water resources. SUMMARY

[0004] The present application provides a dust integrated intelligent management method, device and medium, which is used to solve the following technical problems: the existing stockyard dust management method has hysteresis, and it is difficult to accurately track the spraying management according to the real-time generation of plume dust by the working vehicle, resulting in poor management effect and easy waste of a large amount of water resources.

[0005] The present application adopts the following technical solutions:

[0006] On the one hand, the present application provides a dust integrated intelligent management method, which comprises: performing edge pixel extraction on a target image obtained by a pan-tilt camera to obtain a fitted circumferential trajectory of the target image; combining the fitted circumferential trajectory with the rotation angle of the pan-tilt camera to obtain the visual detection range of the pan-tilt camera; performing target pixel difference detection on the plume dust and the stockyard vehicle in the stockyard through the visual detection range to obtain the three-dimensional coordinate position of the plume dust and the stockyard vehicle; and performing spraying control on a fog gun installed directly above the pan-tilt camera according to the three-dimensional coordinate position, so as to spray the dust reduction area.

[0007] The embodiment of the application monitors the stockyard in real time through the gimbal camera, uses a target detection algorithm to detect plume dust in the stockyard and transport vehicles and excavators that are prone to causing dust in real time, calculates a detected target pixel parallax value, calculates three-dimensional coordinate positions of the detected plume dust and stockyard vehicles in a gimbal coordinate system by using a plume dust parallax model, and locks the plume and stockyard vehicle positions according to the three-dimensional coordinate positions, controls the fog gun to turn, and starts the spraying system to spray, so that the dust production target can be accurately tracked for spraying, the treatment efficiency is high, the effect is good, and a large amount of water resources is saved.

[0008] In a feasible implementation, edge pixels of a target image obtained by a gimbal camera are extracted to obtain a fitting circumferential track of the target image, specifically including: determining a target object in a detection range; obtaining the target image by the gimbal camera, and setting the gimbal camera at this position as an initial position; wherein the target object is located at a center position of the target image; adjusting the gimbal camera to rotate horizontally according to the initial position of the gimbal camera, and obtaining a plurality of test images at different horizontal angles; detecting edge pixels of a gimbal track circle in the plurality of test images, and randomly extracting a plurality of pixel points according to the edge pixels of the gimbal track circle to obtain a fitting circumferential track based on the target image.

[0009] In a feasible implementation, the fitting circumferential track is combined with a rotation angle of the gimbal camera to obtain a visual detection range of the gimbal camera, specifically including: determining a view angle parameter of the gimbal camera according to the rotation angle of the gimbal camera; combining the fitting circumferential track with the rotation angle one by one to obtain a fitting circumferential track parameter; and calculating the visual range of the gimbal camera according to the fitting circumferential track parameter and the view angle parameter to obtain the visual detection range of the gimbal camera.

[0010] The embodiment of the application can calculate the maximum visual range by detecting the visual detection range of the gimbal camera, which provides a guarantee for target recognition and stockyard positioning.

[0011] In a feasible implementation, before the target pixel difference detection of the plume dust and the stockyard vehicle in the stockyard through the visual detection range is performed to obtain the three-dimensional coordinate position of the plume dust and the stockyard vehicle, the method further comprises: collecting a plume dust image in the stockyard; performing back propagation of a residual network on the plume dust image through a preset YOLOX-Darknet backbone neural network structure to obtain a preprocessed plume dust image; obtaining a plurality of video frames corresponding to the preprocessed plume dust image, and performing double-frame difference monitoring on the plurality of video frames to obtain pixel difference change characteristics between the plurality of video frames; performing key supervision training on the YOLOX-Darknet backbone neural network structure according to the pixel change characteristics of the preprocessed plume dust image to obtain a plume dust parallax model; wherein the plume dust parallax model is used to identify the natural protective color of the semi-transparent object and the plume dust.

[0012] The embodiments of the application solve the problems of poor semi-transparent object recognition effect and natural protective color of plume dust by adopting a double-frame difference monitoring frame-to-frame pixel level change scheme and performing key supervision training on the YOLOX-Darknet backbone neural network structure, thereby improving the recognition effect of the plume dust parallax model on the plume dust image.

[0013] In a feasible implementation, after the plume dust parallax model is obtained, the method further comprises: performing vehicle surface light feature extraction on a stockyard vehicle image through a preset deep learning yoloV5+DeepSort algorithm to obtain vehicle light and dark features; performing near neighbor matching on the vehicle light and dark features of the current video frame and the vehicle light and dark features of the adjacent video frame according to the optical flow information of the video frame corresponding to the stockyard vehicle image to perform real-time target tracking on the stockyard vehicle to obtain a target tracking algorithm of the stockyard vehicle.

[0014] The embodiments of the application can effectively improve the poor target tracking effect in the occlusion case by performing near neighbor matching on the vehicle apparent light features extracted in the target vehicle image tracking process, thereby obtaining a target tracking algorithm for real-time tracking of the stockyard vehicle.

[0015] In an implementable embodiment, the target pixel difference detection is performed on the plume dust and the stockyard vehicle in the stockyard through the visual detection range, and the three-dimensional coordinate positions of the plume dust and the stockyard vehicle are obtained, specifically including: reading the RTSP video stream information of the holder camera in real time according to the visual detection range; identifying the pixel position and the pixel height of the corresponding plume dust image features and stockyard vehicle image features in a plurality of RTSP video frames based on the plume dust parallax model and the target tracking algorithm according to the RTSP video stream information, and obtaining real-time target pixel values; wherein the real-time target pixel values include real-time target pixel positions and real-time target pixel heights; performing pixel difference detection calculation on the real-time target pixel values and historical target pixel values to obtain target pixel differences; and judging the three-dimensional positions of the corresponding plume dust image features and stockyard vehicle image features of the target pixel differences through the plume dust parallax model and a preset holder coordinate system, and obtaining the three-dimensional coordinate positions of the plume dust and the three-dimensional coordinate positions of the stockyard vehicle.

[0016] In an implementable embodiment, the spraying control is performed on the fog gun installed directly above the holder camera according to the three-dimensional coordinate positions, so as to spray and reduce dust in the dust reduction area, specifically including: obtaining the three-dimensional coordinate positions of the plume dust and the stockyard vehicle; converting the three-dimensional coordinate positions of the plume dust and the stockyard vehicle into control coordinates, and determining the change area value of the plume dust according to the target pixel differences; wherein the control coordinates are coordinates for controlling the movement of the fog gun; identifying the dust reduction area and the corresponding spraying time according to the change area value of the plume dust based on the control coordinates; and controlling the fog gun to spray and reduce dust according to the dust reduction area and the corresponding spraying time.

[0017] In an implementable embodiment, after the spraying control is performed on the fog gun according to the dust reduction area and the corresponding spraying time, the method further includes: if the spraying time is reached and the target pixel difference is less than a first preset threshold, stopping the control of the fog gun and performing a reset operation on the fog gun installed directly above the holder camera.

[0018] In a second aspect, the embodiments of the present application further provide a dust integrated intelligent management device, which includes: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions capable of being executed by the at least one processor, so that the at least one processor can execute the dust integrated intelligent management method in any of the above embodiments.

[0019] In a third aspect, the embodiments of the present application further provide a non-volatile computer storage medium, characterized in that the storage medium is a non-volatile computer readable storage medium, and the non-volatile computer readable storage medium stores at least one program, each of the programs includes instructions, and the instructions, when executed by a terminal, cause the terminal to perform the dust integrated intelligent management method according to any one of the embodiments.

[0020] The present application provides a dust integrated intelligent management method, device and medium. The material yard is monitored in real time by using a gimbal camera, and a target detection algorithm is used to detect plume dust, transport vehicles and excavators that are prone to cause dust in the material yard in real time. The detected target pixel parallax value is calculated, and the three-dimensional coordinate position of the detected plume dust and material yard vehicle in the gimbal coordinate system is calculated by using a plume dust parallax model. The intelligent linkage system locks the plume and material yard vehicle position according to the three-dimensional coordinate position, controls the fog gun to turn, starts the spraying system to spray, can accurately track the dust production target to spray, has high management efficiency and good effect, and saves a large amount of water resources. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can also be obtained without creative labor. In the drawings:

[0022] Figure 1 A dust integrated intelligent management method flow chart is provided for the embodiments of the present application;

[0023] Figure 2 A corner measurement algorithm analysis flow chart is provided for the embodiments of the present application;

[0024] Figure 3 A multi-target visual detection tracking flow chart is provided for the embodiments of the present application;

[0025] Figure 4 A structure schematic diagram of a dust integrated intelligent management device is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0026] In order for those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0027] The embodiment of the present application provides a dust integrated intelligent management method, as shown in the figure, the dust integrated intelligent management method specifically includes steps S101-S104: Figure 1

[0028] S101, edge pixel extraction is performed on the target image obtained by the cloud platform camera to obtain a fitted circumferential trajectory of the target image.

[0029] Specifically, the target object in the detection range is determined. The target object is imaged by the cloud platform camera to obtain a target image, and the cloud platform camera at this position is set as an initial position. Wherein, the target object is located at the center position of the target image.

[0030] Further, according to the initial position of the cloud platform camera, the horizontal rotation angle of the cloud platform camera is adjusted, and a plurality of test images under different horizontal angles are obtained. The edge pixels of the cloud platform track circle in the plurality of test images are detected, and a plurality of pixel points are randomly extracted based on the edge pixels of the cloud platform track circle to obtain a fitted circumferential trajectory based on the target image.

[0031] In one embodiment, the camera with the cloud platform is fixed directly above the fog gun, the cloud platform automatically adjusts the rotation angle of the camera, so that the target object in the camera image is located at the image center, and this camera angle is fixed. Then adjust the rotation angle of the camera to obtain a plurality of test images under different horizontal rotation angles. The edge pixels of the cloud platform track circle are detected by using the image edge detection method, and a plurality of pixel points on the edge of the track circle are randomly extracted, and then all the pixel points are fitted as a circumferential trajectory.

[0032] S102, the fitted circumferential trajectory is combined with the rotation angle of the cloud platform camera to obtain the visual detection range of the cloud platform camera.

[0033] Specifically, according to the rotation angle of the cloud platform camera, the view angle parameters of the cloud platform camera are determined. The fitted circumferential trajectory is combined with the rotation angle one by one to obtain the fitted circumferential trajectory parameters. According to the fitted circumferential trajectory parameters and the view angle parameters, the visual range of the cloud platform camera is calculated to obtain the visual detection range of the cloud platform camera.

[0034] In one embodiment, Figure 2 ​An angle measurement algorithm analysis flowchart provided for an embodiment of the present application is shown in Figure 2 As shown, before obtaining the target height parameter, it is further judged whether there is enough test image information for corner correction, until enough test images are obtained, then the rotation angle of the holder camera is determined to determine the view angle parameter of the holder camera, and then the fitting circumferential trajectory parameter is combined to obtain the fitting circumferential trajectory parameter, and then the visual range of the holder camera is calculated according to the obtained target height parameter of the target object, and finally the visual detection range of the holder camera is determined.

[0035] S103, through the visual detection range, target pixel difference detection of plume dust and stockyard vehicles in the stockyard is performed to obtain the three-dimensional coordinate positions of the plume dust and the stockyard vehicles.

[0036] Specifically, the plume dust image in the stockyard is collected. The pre-processed plume dust image is obtained by using the residual network back propagation method of the pre-set YOLOX-Darknet backbone neural network structure. A plurality of video frames corresponding to the pre-processed plume dust image are obtained, and the pixel difference change features between the plurality of video frames are obtained by double-frame difference monitoring.

[0037] Further, the backbone neural network structure of YOLOX-Darknet is supervised and trained according to the pixel change features of the pre-processed plume dust image, and a plume dust parallax model is obtained. The plume dust parallax model is used to train and identify semi-transparent objects and natural protective colors of plume dust.

[0038] In one embodiment, by using a target detection model of a deep learning convolutional neural network, the model is trained by collecting the plume dust image of the implementation site, the backbone neural network structure of YOLOX-Darknet is selected as the model selection, the residual network back propagation method is used to reduce the gradient disappearance method for visual detection of plume dust, which can meet the high-performance detection requirement, without pre-setting the size of the prior box, which facilitates sufficient learning of small objects, then the double-frame difference is used to monitor the pixel-level pixel difference change features between frames, and the network structure is modified for supervised training to obtain a plume dust parallax model to solve the problems of poor semi-transparent object recognition and natural protective color of plume dust.

[0039] Meanwhile, by presetting a deep learning yoloV5+DeepSort algorithm, vehicle surface light features of the stockyard vehicle image are extracted to obtain vehicle light and dark features. According to optical flow information of the stockyard vehicle image corresponding to a video frame, the vehicle light and dark features of the current video frame are matched with the vehicle light and dark features of the adjacent video frame to track the stockyard vehicle in real time, and a target tracking algorithm of the stockyard vehicle is obtained.

[0040] In one embodiment, Figure 3 A multi-target visual detection tracking flowchart provided by the embodiment of the application is shown in Figure 3 As shown, a deep learning yoloV5+DeepSort algorithm is used to track and count pedestrians and vehicles in a video. DeepSort extracts and filters the deep apparent features of the observed target, and performs Kalman filtering and corresponding feature matching. Then, according to the optical flow information between video frames, the apparent light and dark features of the target extracted in the target tracking process are matched between adjacent frames. In this way, the target tracking effect under the condition of occlusion can be effectively improved. Finally, a target tracking algorithm that can track the stockyard vehicle in real time is obtained, and the plume dust caused by the stockyard vehicle is detected in real time.

[0041] Further, according to the visual detection range, RTSP video stream information of the pan-tilt camera is read in real time. According to the RTSP video stream information, based on a plume dust parallax model and a target tracking algorithm, the corresponding plume dust image features and stockyard vehicle image features in a plurality of RTSP video frames are identified in terms of pixel position and pixel height to obtain real-time target pixel values. The real-time target pixel values include real-time target pixel positions and real-time target pixel heights.

[0042] Further, the real-time target pixel values and historical target pixel values are detected and calculated in terms of pixel difference to obtain target pixel differences. By the plume dust parallax model and a preset pan-tilt coordinate system, the plume dust image features and stockyard vehicle image features corresponding to the target pixel differences are judged in terms of three-dimensional position to obtain three-dimensional coordinate positions of the plume dust and three-dimensional coordinate positions of the stockyard vehicle.

[0043] S104, according to the three-dimensional coordinate positions, a fog gun installed directly above the pan-tilt camera is controlled to spray to spray and reduce dust in the dust reduction area.

[0044] Specifically, the three-dimensional coordinate positions of the plume dust and the stockyard vehicle are obtained. The three-dimensional coordinate positions of the plume dust and the stockyard vehicle are converted into control coordinates, and the change area value of the plume dust is determined according to the target pixel differences. The control coordinates are coordinates for controlling the movement of the fog gun.

[0045] Further, based on the control coordinates, a to-be-dust-settling region and corresponding spraying time are identified according to the changing area value of the plume dust. The fog gun is controlled to spray dust settling according to the to-be-dust-settling region and the corresponding spraying time. If the spraying time is reached and the target pixel difference is less than the first preset threshold, the fog gun is stopped and the fog gun installed directly above the gimbal camera is reset.

[0046] In one embodiment, the dust integrated intelligent control system is composed of a field monitoring AI sensing system, an intelligent linkage control system and an integrated management platform. The field monitoring AI sensing system, the intelligent linkage control system and the integrated management platform are composed. The AI sensing system includes a gimbal camera and a plume dust parallax model. The intelligent linkage control system includes a gimbal control module, a fog gun control module and a spraying control module. The integrated management platform includes real-time monitoring of the stockyard, data monitoring of the fog gun and the camera, etc. First, the field monitoring AI sensing system is used to monitor the stockyard in real time. According to the monitored data, the integrated management platform controls the intelligent linkage control system to treat the to-be-dust-settling region, and finally realizes intelligent control of dust.

[0047] In one embodiment, the plume dust generated by the transport vehicle in the field of view of the gimbal camera and the plume dust generated in some areas are identified. The AI visual sensing system performs real-time target detection to detect the specific three-dimensional coordinate position of the substandard plume dust region, and then starts the fog gun to convert the specific three-dimensional coordinate of the substandard plume dust region into control coordinates of the fog gun, controls the fog gun to serve the dust-settling region to implement spraying dust settling, and when the fog gun continues for a certain time and the corresponding target pixel difference of the smoke dust region is less than the first preset threshold, it is considered that the to-be-dust-settling region has been successfully sprayed and settled, the fog gun stops spraying and settling, and finally all devices are reset.

[0048] In addition, the present application also provides a dust integrated intelligent control device, as shown in Figure 4 The dust integrated intelligent control device 400 specifically includes:

[0049] at least one processor 401; and a memory 402 connected with the at least one processor 401; wherein the memory 402 stores instructions executable by the at least one processor 401, so that the at least one processor 401 can execute:

[0050] performing edge pixel extraction on the target image obtained by the gimbal camera to obtain a fitted circumferential trajectory of the target image;

[0051] combining the fitted circumferential trajectory with the rotation angle of the gimbal camera to obtain a visual detection range of the gimbal camera;

[0052] The target pixel difference of the plume dust and the stockyard vehicle in the stockyard is detected through the visual detection range, and the three-dimensional coordinate position of the plume dust and the stockyard vehicle is obtained;

[0053] According to the three-dimensional coordinate position, the fog gun installed above the pan-tilt camera is controlled to spray, so as to spray the dust falling area.

[0054] The embodiment of the present application provides a dust integrated intelligent treatment method and device, and medium. The stockyard is monitored in real time through the pan-tilt camera, and the plume dust, the transport vehicle and the excavator which are easy to cause dust in the stockyard are detected in real time by using a target detection algorithm. The target pixel parallax value of the detected target is calculated, and the three-dimensional coordinate position of the detected plume dust and stockyard vehicle in the pan-tilt coordinate system is calculated by using a plume dust parallax model. The intelligent linkage system locks the plume and the stockyard vehicle position according to the three-dimensional coordinate position, controls the fog gun to turn, and starts the spraying system to spray. The dust production target can be accurately tracked for spraying, the treatment efficiency is high, the effect is good, and a large amount of water resources can be saved.

[0055] Each of the embodiments in the present application is described in a progressive manner, and the same and similar parts of each of the embodiments can be referred to each other. Each of the embodiments mainly describes the difference from other embodiments. Especially, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple, and the related parts can be referred to the part of the method embodiment.

[0056] The device and medium provided by the embodiment of the present application are one-to-one corresponding to the method, so the device and medium also have the similar beneficial technical effects of the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here.

[0057] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0058] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions specified in the flowchart block or blocks. Figure 1 The flowchart and / or block diagram in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to the present disclosure. In this regard, each flowchart block and / or block in the Figures can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that the flowchart blocks and / or blocks in the Figures can represent a procedure, apparatus, and / or device with features enabling the performance of the functions specified in the flowchart block and / or blocks and combinations of the flowchart block and / or blocks with features enabling the performance of the functions specified in the flowchart block and / or blocks specified therein. The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions specified in the flowchart block or blocks. Figure 1 The flowchart and / or block diagram in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to the present disclosure. In this regard, each flowchart block and / or block in the Figures can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that the flowchart blocks and / or blocks in the Figures can represent a procedure, apparatus, and / or device with features enabling the performance of the functions specified in the flowchart block and / or blocks and combinations of the flowchart block and / or blocks with features enabling the performance of the functions specified in the flowchart block and / or blocks specified therein. The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions specified in the flowchart block or blocks.

[0059] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, etc. The memory is an example of computer readable media.

[0060] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0061] It should also be noted that the terms "comprising", "containing", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements in the list, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0062] The above merely provides an example of the present application, but is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the scope of claims of the present application.

Claims

1. A dust integrated intelligent management method, characterized in that, The method comprises: Edge pixel extraction is performed on a target image obtained by a pan-tilt camera to obtain a fitted circumferential track of the target image; The fitted circumferential track is combined with a rotation angle of the pan-tilt camera to obtain a visual detection range of the pan-tilt camera; An image of plume dust in a stockyard is collected; A pre-processing plume dust image is obtained by performing a residual network back propagation on the image of plume dust through a preset YOLOX-Darknet backbone neural network structure; A plurality of video frames corresponding to the pre-processing plume dust image are obtained, and a pixel difference change feature between the plurality of video frames is obtained by performing double-frame difference monitoring on the plurality of video frames; A plume dust parallax model is obtained by performing key supervision training on the YOLOX-Darknet backbone neural network structure according to the pixel change feature of the pre-processing plume dust image, wherein the plume dust parallax model is used to identify a semi-transparent object and a natural protective color of plume dust; A vehicle surface light feature of a stockyard vehicle image is extracted through a preset deep learning yoloV5+DeepSort algorithm to obtain a vehicle bright-dark feature; The vehicle bright-dark feature of a current video frame is matched with a vehicle bright-dark feature of a neighboring video frame according to light flow information of a video frame corresponding to the stockyard vehicle image, so as to perform real-time target tracking on the stockyard vehicle and obtain a target tracking algorithm of the stockyard vehicle; Target pixel difference detection is performed on plume dust and a stockyard vehicle in a stockyard through the visual detection range to obtain three-dimensional coordinate positions of the plume dust and the stockyard vehicle, specifically including: Real-time RTSP video stream information of the pan-tilt camera is read according to the visual detection range; Pixel positions and pixel heights of corresponding plume dust image features and stockyard vehicle image features in a plurality of RTSP video frames are identified based on the plume dust parallax model and the target tracking algorithm according to the RTSP video stream information, to obtain real-time target pixel values; wherein the real-time target pixel values include real-time target pixel positions and real-time target pixel heights; Target pixel difference is obtained by performing pixel difference detection calculation on the real-time target pixel values and historical target pixel values; Three-dimensional positions of the plume dust image features and the stockyard vehicle image features corresponding to the target pixel difference are judged through the plume dust parallax model and a preset pan-tilt coordinate system to obtain three-dimensional coordinate positions of the plume dust and three-dimensional coordinate positions of the stockyard vehicle; A fog gun installed directly above the pan-tilt camera is controlled to spray and dust a dust reduction area according to the three-dimensional coordinate positions.

2. The dust integrated intelligent management method according to claim 1, characterized in that, Edge pixel extraction is performed on a target image obtained by a pan-tilt camera to obtain a fitted circumferential track of the target image, specifically including: A target object in a detection range is determined; The gimbal camera is used to acquire an image of the target object to obtain the target image, and the gimbal camera at this position is set as an initial position; wherein the target object is located at the center of the target image; According to the initial position of the gimbal camera, the horizontal rotation angle of the gimbal camera is adjusted, and a plurality of test images at different horizontal angles are acquired; The edge pixels of the gimbal track circle in the plurality of test images are detected, and a plurality of pixel points are randomly extracted based on the edge pixels of the gimbal track circle to obtain a fitted circumferential track based on the target image.

3. The dust integrated intelligent management method according to claim 1, characterized in that, The fitted circumferential track and the rotation angle of the gimbal camera are combined to obtain the visual detection range of the gimbal camera, specifically including: According to the rotation angle of the gimbal camera, the view angle parameters of the gimbal camera are determined; The fitted circumferential track and the rotation angle are one-to-one combined to obtain the fitted circumferential track parameters; According to the fitted circumferential track parameters and the view angle parameters, the visual range of the gimbal camera is calculated to obtain the visual detection range of the gimbal camera.

4. The dust integrated intelligent management method according to claim 1, characterized in that, According to the three-dimensional coordinate position, the fog gun installed directly above the gimbal camera is controlled to spray to reduce dust in the dust area, specifically including: The three-dimensional coordinate positions of the smoke dust and the stockyard vehicle are acquired; The three-dimensional coordinate positions of the smoke dust and the stockyard vehicle are converted into control coordinates, and the change area value of the smoke dust is determined according to the target pixel difference; wherein the control coordinates are the coordinates for controlling the movement of the fog gun; Based on the control coordinates, the change area value of the smoke dust is identified to obtain the dust area to be reduced and the corresponding spraying time; According to the dust area to be reduced and the corresponding spraying time, the fog gun is controlled to spray dust.

5. The dust integrated intelligent management method according to claim 4, characterized in that, After controlling the fog gun to spray dust according to the dust area to be reduced and the corresponding spraying time, the method further includes: If the spraying time is reached and the target pixel difference is less than a first preset threshold, the fog gun is stopped and the fog gun installed directly above the gimbal camera is reset.

6. A dust integrated intelligent management device, characterized in that, The device includes: At least one processor; and The memory is in communication with the at least one processor; wherein The memory stores instructions executable by the at least one processor to enable the at least one processor to execute the dust integrated intelligent management method according to any one of claims 1-5.

7. A non-transitory computer storage medium, comprising, The storage medium is a non-volatile computer readable storage medium, and the non-volatile computer readable storage medium stores at least one program, and each program includes instructions, which when executed by a terminal, causes the terminal to execute the dust integrated intelligent management method according to any one of claims 1-5.

Images