Dust detection methods, devices, computer equipment and storage media

Abstract

This application relates to a dust detection method, apparatus, computer equipment, and storage medium. By acquiring initial image data containing the vehicle body and cargo after the vehicle enters the detection area, accurately identifying the boundaries of the carriages and the gaps between the couplings, and completing the image segmentation of a single carriage, the dust suppressant coverage detection is performed on each carriage and the carriage identification is bound. This upgrades the traditional coarse-grained detection of the entire train to fine-grained detection of a single carriage, quickly locating the position of unqualified carriages, and realizing a one-to-one correspondence between the detection results and the carriage information. This solves the problems of traditional detection being unable to accurately trace, difficult to locate, and inefficient.

Description

Technical Field

[0001] This application relates to the field of coal transportation monitoring technology, and in particular to a dust detection method, apparatus, computer equipment, and storage medium. Background Technology

[0002] In the railway freight system, especially in the heavy-haul coal transportation process, coal dust has always been a key focus of environmental protection supervision.

[0003] Traditional technologies typically only detect coal dust emissions from trains transporting coal, which presents challenges such as high particle size and difficulty in locating vehicles that fail the coal dust emission test. Summary of the Invention

[0004] Therefore, it is necessary to provide a dust detection method, device, computer equipment, and storage medium to address the aforementioned technical problems, which can quickly locate the position of vehicles that fail to meet coal dust emission standards.

[0005] In a first aspect, this application provides a dust detection method, comprising:

[0006] When a transport vehicle enters the detection area, at least one frame of initial image data of the transport vehicle is acquired; each frame of initial image data includes vehicle body data and cargo data.

[0007] Based on the vehicle body data, identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments;

[0008] Based on the identified car boundaries and the gap between different cars, each frame of initial image data is segmented to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0009] For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

[0010] In one embodiment, identifying the boundaries of each compartment of the transport vehicle and the hook gap area between different compartments based on the vehicle body data includes:

[0011] For each initial image data, the initial image data is processed to obtain the semantic mask features corresponding to the initial image data; wherein, the semantic mask features include at least the mask features of the carriage and the gap between the couplers;

[0012] Feature analysis is performed on each of the semantic mask features to obtain the boundary of each compartment of the transport vehicle and the hook gap area between different compartments.

[0013] In one embodiment, processing the initial image data to obtain the semantic mask features corresponding to the initial image data includes:

[0014] Extract color image features from the initial image data; wherein, the color image features include color and texture information of the carriage and the gap between the couplers;

[0015] The initial image data is processed to grayscale and gradient is calculated to obtain a gradient map containing gradient magnitudes.

[0016] The gradient map is subjected to coordinate transformation and depth calculation to obtain a depth feature map;

[0017] The gradient map and the depth feature map are fused to obtain a fused feature map;

[0018] The fused feature map is input into a semantic segmentation model for semantic recognition to obtain the semantic mask features of the initial image data.

[0019] In one embodiment, the step of performing feature analysis on each of the semantic mask features to obtain the boundary of each compartment of the transport vehicle and the hook gap region between different compartments includes:

[0020] Based on the mask features of the carriages in each of the semantic mask features, determine the outermost contour of each carriage;

[0021] The boundary of each compartment of the transport vehicle is determined based on the outermost contour of each compartment.

[0022] Connectivity analysis is performed on the mask features of the hook-and-gear gaps between different carriages in each of the semantic mask features to obtain the hook-and-gear gap regions between different carriages of the transport vehicle.

[0023] In one embodiment, the initial image data of each frame is segmented based on the identified carriage boundaries and the gap area between different carriages to obtain multiple segmented image data, including:

[0024] Using the gap between the car bodies as the car body boundary, each frame of the initial image data is segmented based on the identified car body boundary to obtain multiple segmented image data.

[0025] In one embodiment, the step of performing dust detection on the cargo data in the segmented image data to obtain dust detection results including inhibitor detection results and carriage markings includes:

[0026] Based on the cargo data in the segmented image data, determine the cargo area and the uncovered area of ​​the cargo surface without dust inhibitor;

[0027] The inhibitor detection results are determined based on the cargo area and the uncovered area;

[0028] Based on the inhibitor detection results and the carriage identification corresponding to the segmented image data, dust detection results are generated.

[0029] In one embodiment, determining the inhibitor detection result based on the cargo area and the uncovered area includes:

[0030] If the uncovered area is the total uncovered area, and the ratio of the total uncovered area to the cargo area is greater than or equal to a first threshold, then the inhibitor test result is determined to be unqualified; or,

[0031] If the uncovered area includes the independent uncovered area of ​​each dispersed region, and there exists an independent uncovered area that is greater than or equal to the second threshold, then the dust detection result is determined to be unqualified; wherein, the first threshold is greater than the second threshold.

[0032] Secondly, this application also provides a dust detection device, comprising:

[0033] The acquisition module is used to acquire at least one frame of initial image data of the transport vehicle when the transport vehicle enters the detection area; each frame of initial image data includes vehicle body data and cargo data.

[0034] The identification module is used to identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments based on the vehicle body data.

[0035] The segmentation module is used to segment each frame of initial image data acquired based on the identified car boundaries and the gap area between different cars, to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0036] The detection module is used to perform dust detection on the cargo data in each segmented image data, and obtain dust detection results including inhibitor detection results and carriage markings.

[0037] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0038] When a transport vehicle enters the detection area, at least one frame of initial image data of the transport vehicle is acquired; each frame of initial image data includes vehicle body data and cargo data.

[0039] Based on the vehicle body data, identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments;

[0040] Based on the identified car boundaries and the gap between different cars, each frame of initial image data is segmented to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0041] For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

[0042] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0043] When a transport vehicle enters the detection area, at least one frame of initial image data of the transport vehicle is acquired; each frame of initial image data includes vehicle body data and cargo data.

[0044] Based on the vehicle body data, identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments;

[0045] Based on the identified car boundaries and the gap between different cars, each frame of initial image data is segmented to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0046] For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

[0047] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0048] When a transport vehicle enters the detection area, at least one frame of initial image data of the transport vehicle is acquired; each frame of initial image data includes vehicle body data and cargo data.

[0049] Based on the vehicle body data, identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments;

[0050] Based on the identified car boundaries and the gap between different cars, each frame of initial image data is segmented to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0051] For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

[0052] The aforementioned dust detection method, device, computer equipment, and storage medium acquire initial image data containing the vehicle body and cargo after the vehicle enters the detection area, accurately identify the boundaries of the carriages and the gaps between the couplings, and complete the image segmentation of a single carriage. Then, dust suppressant coverage detection is performed on each carriage and carriage identification is bound. This upgrades the traditional coarse-grained detection of the entire train to fine-grained detection of a single carriage, quickly locates the position of unqualified carriages, and achieves a one-to-one correspondence between the detection results and carriage information. This solves the problems of traditional detection, such as inaccurate traceability, difficulty in positioning, and low efficiency. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a diagram illustrating the application environment of a dust detection method in one embodiment;

[0055] Figure 2 This is a flowchart illustrating a dust detection method in one embodiment;

[0056] Figure 3 This is a flowchart illustrating the process of identifying the car boundary and the gap area between the coupler in one embodiment.

[0057] Figure 4 This is a schematic diagram illustrating the process of obtaining semantic mask features of initial image data in one embodiment;

[0058] Figure 5 This is a schematic diagram of the process for obtaining dust detection results in one embodiment;

[0059] Figure 6 This is a structural block diagram of a dust detection device in one embodiment;

[0060] Figure 7This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0062] The dust detection method provided in this application embodiment can be applied to, for example... Figure 1 In the illustrated application environment, for example at tunnel entrances and exits, dust detection is performed on carriages transporting heavy-duty goods (e.g., coal). This method can be executed by an industrial control computer (ICC) within the dust detection system. The ICC can be a server or a terminal, or a server and a terminal; for example, it can be an embedded ICC (edge ​​computing terminal) and a local real-time computing server. The dust detection system also includes:

[0063] The top scanning unit (e.g., a camera) is mounted on a cantilever bracket on the tunnel arch and takes pictures vertically downwards from directly above the carriage; the camera's field of view can completely cover the cargo area on top of the carriage; the installation height can be approximately 5.5–8 meters above the rail surface and approximately 5–8 meters above the top surface of the carriage.

[0064] Side / vehicle number recognition unit (e.g., camera), a device installed on the tunnel sidewall, to photograph the carriage area from the side: the field of view of the side camera is used to supplement the photograph of the side of the carriage and identify the vehicle number.

[0065] The top scanning unit and the side / vehicle number recognition unit can use a dedicated intelligent area array scanning camera for vehicle scanning systems, which is smaller, more efficient, and has higher image fidelity, while also embedding intelligent analysis functions; the built-in image signal processor speed detection (ISP) technology does not have strict requirements for the vehicle's speed, supporting the scanning and detection of vehicles with speeds up to 60km / h; based on the independently developed dedicated image acquisition unit for scanning systems and the specially designed application-specific integrated circuit (ASIC) processor, it can achieve high-definition color scanning images with a resolution of 4096p.

[0066] Optional technical parameters for the dust detection system include power supply 100-240VAC, operating temperature -30°~+70°, imaging component protection rating IP66, lighting component protection rating IP66, lens field of view ≥100°, and image data interface Gigabit Ethernet.

[0067] System Functions: Supports multiple vehicle types, suitable for open wagons, boxcars, tank cars, flatbed wagons, etc.; features intelligent vehicle number recognition, utilizing image acquisition and intelligent recognition technologies to intelligently identify basic information such as vehicle numbers; possesses rich image post-processing (ISP) functions, capable of performing various enhancement operations on scanned images, such as brightness, saturation, contrast, white balance, and sharpening; supports high-speed vehicle passage for image output.

[0068] Once the train enters the tunnel detection area, data acquisition is triggered. The top scanning unit takes a top-down view of the vehicle and cargo to capture the entire scene for suppressor coverage and dust detection; the side / car number recognition unit captures images of the sides of the carriages for car number recognition and posture correction; the captured image data is transmitted back to the industrial control computer in real time for dust detection.

[0069] In one exemplary embodiment, such as Figure 2 As shown, a dust detection method is provided. Taking the application of this method to an industrial control computer as an example, the method includes the following steps:

[0070] S201, when the transport vehicle enters the detection area, acquire at least one frame of initial image data of the transport vehicle.

[0071] Among them, transport vehicles refer to train formations used for transporting bulk materials such as coal and ore on heavy-haul railway lines, which are composed of multiple independent carriages connected by couplers.

[0072] The detection area refers to a fixed monitoring section designated at locations such as the entrance and exit of railway tunnels and the exit of dust suppressant spraying stations. It is equipped with a complete set of equipment, including image acquisition, trigger sensing, and supplementary lighting. The physical detection and image acquisition process is initiated as soon as a vehicle enters the area.

[0073] Initial image data refers to one or more frames of color digital images acquired by a high-definition intelligent area array camera after the transport vehicle enters the detection area and triggers the acquisition. Each frame of initial image data includes vehicle body data and cargo data, that is, the image contains complete information on the vehicle body structure and the cargo loaded in the vehicle, and is the original data for all subsequent recognition, segmentation and detection steps.

[0074] Car body data refers to all structural information related to the train carriages in the initial image, including geometric and textural feature data such as the carriage outline, carriage side panels, carriage ends, connecting parts between carriages, and coupler gaps.

[0075] Cargo data refers to image data related to the condition of the coal loaded inside the wagon in the initial image, such as surface texture, color distribution, stacking shape, and dust suppressant coverage traces.

[0076] For example, when the axle of the transport vehicle triggers the laser beam sensor in the detection area, the top high-definition area array camera and the side vehicle number camera are activated, and the high-brightness supplementary light is turned on simultaneously to eliminate the influence of tunnel shadows and low light at night. Multiple frames of images are continuously acquired and images without motion blur and overexposure are selected as the initial image data to ensure that clear vehicle structure and cargo surface information are included at the same time.

[0077] S202 identifies the boundaries of each compartment of the transport vehicle and the gap area between different compartments based on the vehicle body data.

[0078] The carriage boundary refers to the outer closed contour range of the area occupied by a single carriage in the image coordinate system. It can be composed of the coordinate points of the upper left, upper right, lower left, and lower right, and is used to clearly distinguish the independent image range of each carriage.

[0079] The hook gap area refers to the gap area between the ends of two adjacent carriages where the coupler is connected. This area is not covered by cargo, has a clearly interrupted texture, and its grayscale and color are significantly different from the main body of the carriage.

[0080] For example, feature enhancement, edge extraction, and semantic analysis can be performed on the vehicle body data in the initial image data to locate all the main areas of the carriages. Then, the hook gap between adjacent carriages can be identified through contour features and spacing features, and finally, the boundary coordinates and gap area coordinates of each carriage can be output.

[0081] S203, based on the identified car boundaries and the gap area between different cars, perform image segmentation on each frame of initial image data to obtain multiple segmented image data.

[0082] Each segmented image data is an image of a single carriage, and the segmented image data includes the cargo data transported by the carriage. That is, the segmented image data refers to the independent image of a single carriage obtained by cropping according to the carriage boundary with the gap between the couplers as the dividing line. Each image contains only one carriage and its internal cargo, which is used to achieve independent detection of "one carriage, one image".

[0083] For example, the dividing line between the carriages can be determined based on the gap between the couplers, and combined with the identified carriage boundary coordinates, the initial image can be precisely cropped and separated, splitting the multiple initial image data corresponding to the entire train into multiple independent single carriage images. Each segmented image completely retains the car body and cargo information of the corresponding carriage, without loss, overlap, or truncation.

[0084] In some alternative implementations, the gap between the hooks between the carriages can be used as the carriage boundary. Based on the identified carriage boundaries, each frame of the initial image data is segmented to obtain multiple segmented image data.

[0085] For example, the initial image data contains 3 consecutive carriages and 2 gaps between the couplings; the two gaps can be used as the dividing lines between the carriages, and the entire image can be cropped into 3 independent segmented images along the identified carriage boundaries. Each image contains only the complete car body and cargo data of one carriage, with no background redundancy and no interference from adjacent carriages.

[0086] In this way, image segmentation is performed using the gap between the hooks as the boundary, ensuring that the segmented images strictly correspond to a single carriage. This achieves true one-car-one-image precision segmentation, avoiding problems such as carriage adhesion, truncation, and overlap, and improving the accuracy and consistency of subsequent dust suppressant detection.

[0087] S204. For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results including inhibitor detection results and carriage markings.

[0088] Dust detection refers to an automated detection process based on machine vision and image analysis to identify, calculate, and determine the uniformity and integrity of dust suppression inhibitor spraying on the surface of goods, with the aim of determining whether dust pollution will occur.

[0089] The inhibitor test result refers to the qualified or unqualified judgment output after the dust test is completed, which is used to characterize whether the dust suppressant meets the environmental protection coverage requirements.

[0090] Carriage identification refers to information such as carriage number, carriage number, and formation order that can uniquely identify a single carriage, used to achieve accurate location and full traceability of unqualified carriages.

[0091] For example, cargo area extraction, dust suppressant coverage area calculation, and bare area determination can be performed on each single carriage segmentation image. A qualified or unqualified conclusion can be obtained according to preset rules, and the conclusion can be bound to the unique identifier of the carriage to form a queryable and traceable dust detection result.

[0092] Taking the inspection of a heavy-haul coal train as an example, assuming the train consists of 3 open wagons and enters the tunnel inspection area at 25 km / h, the initial image with a resolution of 4096×2160 is acquired by the acquisition equipment, containing complete data on the wagon body and coal cargo. Then, based on the wagon body data, the rectangular boundaries of the 3 wagons and the two gaps between the wagons are identified. Using these two gaps as boundaries, the initial image is segmented into 3 independent single-wagon segment images. Dust suppressant coverage is detected on each of the 3 segment images, and the final inspection results are output with wagon numbers. For example, wagon C64K-12345 is qualified; wagon C64K-12346 is unqualified; and wagon C64K-12347 is qualified.

[0093] In the above embodiments, by acquiring initial image data containing the vehicle body and cargo after the vehicle enters the detection area, accurately identifying the boundaries of the carriages and the gap between the couplers, and completing the image segmentation of a single carriage, the dust suppressant coverage detection is carried out on each carriage and the carriage identification is bound. This can upgrade the traditional coarse-grained detection of the entire train to fine-grained detection of a single carriage, quickly locate the position of the unqualified carriage, and realize the one-to-one correspondence between the detection results and the carriage information. This solves the problems of traditional detection being unable to accurately trace, difficult to locate, and inefficient.

[0094] In some alternative implementations, see [link to relevant documentation]. Figure 3 , Figure 3 A flowchart illustrating the process for identifying the boundaries of a car body and the gap area between the couplers is provided, specifically including the following steps:

[0095] S301, For each initial image data, process the initial image data to obtain the semantic mask features corresponding to the initial image data.

[0096] The semantic mask features include at least the mask features of the carriage and the gap between the couplers. That is, the semantic mask features refer to the feature map generated by pixel-level semantic classification processing, which uses different pixel values ​​or color channels to mark different objects such as "carriage area", "coupler gap area" and "background area", and at least the mask features of the carriage and the gap between the couplers.

[0097] For example, each initial image data can be processed by multi-feature extraction, image enhancement and normalization, and then the processed feature data can be input into the trained semantic segmentation model. The model can classify each pixel and output semantic mask features that respectively label the carriage, the gap between the couplers and the background. Different categories are distinguished by different pixel values ​​and do not interfere with each other.

[0098] In some optional implementations, the outermost contour of each car can be determined based on the mask features of the car in each semantic mask feature; then, based on the outermost contour of each car, the boundary of each car of the transport vehicle can be determined; and connected component analysis can be performed on the mask features of the hook gap between different cars in each semantic mask feature to obtain the hook gap region between different cars of the transport vehicle.

[0099] The outermost contour refers to the outermost closed polygonal contour of the carriage mask area, excluding internal contours such as holes, recesses, and cargo obstructions inside the carriage, and is used to accurately define the overall scope of the carriage.

[0100] Connected component analysis refers to the analysis method of detecting, marking, calculating the area, and statistically analyzing the bounding rectangles of regions with the same pixel value that are interconnected in an image. It is used to filter out real and valid gaps from the mask and eliminate false gaps such as noise, stains, and shadows.

[0101] For example, the carriage mask can be binarized and morphologically optimized, all closed contours can be extracted using a contour retrieval algorithm, and only the outermost contour can be retained, while internal holes and interference can be filtered out to obtain the outermost contour of each carriage.

[0102] Furthermore, polygon fitting and circumscribed rectangle calculation can be performed on the outermost contour to obtain the coordinates of the upper left, upper right, lower left, and lower right of the carriage, forming the standard carriage boundary for subsequent image segmentation.

[0103] Furthermore, the gap mask can be binarized and morphologically processed, then connected component detection and filtering can be performed. Effective gaps can be retained based on physical width, aspect ratio, and positional features, while interference areas can be eliminated, ultimately yielding accurate gap coordinates.

[0104] For example, the outermost rectangular outline of each car is extracted based on the car mask features. The car boundary coordinates are calculated from the outline as [(120, 60), (1250, 60), (1250, 820), (120, 820)]. Connected component analysis is performed on the hook-and-gear gap mask to select effective areas with a physical width of about 10cm that conform to the real gap features. The center coordinates and left and right boundary coordinates of the gap are output, and finally the accurate car boundary and hook-and-gear gap area are obtained.

[0105] In this way, by extracting the outermost contour of the carriage to determine the boundary and filtering the effective hook gap through connected component analysis, the range of a single carriage can be accurately divided. At the same time, the interference of false gaps, noise, and stains can be eliminated, making the segmentation boundary more stable and reliable, and providing solid support for subsequent accurate segmentation of one carriage per map.

[0106] S302, perform feature analysis on each semantic mask feature to obtain the boundary of each compartment of the transport vehicle and the hook gap area between different compartments.

[0107] Then, the semantic mask features are analyzed by contour extraction, connected component statistics, and geometric filtering. The boundary contour is extracted from the car body mask, and the effective gap area is located from the gap mask. Finally, a stable and reliable car body boundary and coupler gap position are output.

[0108] For example, after performing multi-dimensional feature processing on the initial image data of a coal train, a semantic mask feature map is generated. In the semantic mask feature map, the pixel value of the car area is marked as 1, the pixel value of the coupler gap area is marked as 2, and the background area is marked as 0. Contour analysis is performed on the area with a pixel value of 1 to obtain the boundary of each car; region analysis is performed on the area with a pixel value of 2 to obtain the position of the coupler gap between adjacent cars, and finally, a complete set of boundary and gap coordinates is output.

[0109] In the above embodiments, the initial image data is first processed to generate semantic mask features containing the gap between the carriage and the coupler, and then the mask features are subjected to structured analysis. This enables stable differentiation between the main body of the carriage and the gap area in complex on-site environments such as dust, shadows, and uneven lighting, thereby improving the recognition accuracy and anti-interference ability.

[0110] In some alternative implementations, see [link to relevant documentation]. Figure 4 , Figure 4 A flowchart illustrating the process of obtaining semantic mask features from initial image data is provided, specifically including the following steps:

[0111] S401, Extract color image features from the initial image data.

[0112] Among them, color image features include color and texture information of the carriage and the gap between the coupler; color image features refer to the color distribution, texture details, brightness changes and other features extracted from the initial image data (RGB color image), which can reflect the appearance difference between the carriage and the gap between the coupler and improve semantic distinguishability.

[0113] For example, color histograms, texture gradients, and local features can be extracted from the initial image data to preserve the differences between the metal texture of the carriage and the dark areas in the gaps. For instance, color space normalization can be performed on the initial image data, mapping the R / G / B channel pixel values ​​to the 0-1 range to eliminate the influence of varying lighting intensity. Color histogram features can then be extracted; for example, the color distribution of the carriage's metal areas, the dark areas in the hook gaps, and the cargo areas can be statistically analyzed separately. Furthermore, the Local Binary Pattern (LBP) operator or the difference of Gaussian operator can be used to obtain the texture differences between the carriage surface texture and the gap areas. Finally, the color and texture features are concatenated to obtain the color image features.

[0114] S402, perform grayscale processing and gradient calculation on the initial image data to obtain a gradient map containing gradient magnitude.

[0115] A gradient map is an image obtained by calculating the pixel gradient magnitude after converting the image to grayscale. It is used to highlight structural information such as edges, contours, and corners, and to enhance the boundary features between carriages and gaps.

[0116] For example, a color image can be converted to a grayscale image, and the horizontal and vertical gradients can be calculated using the Sobel or Canny operator to synthesize a gradient magnitude map, making the edges and gaps of the carriage more prominent.

[0117] S403 performs coordinate transformation and depth calculation on the gradient map to obtain a depth feature map.

[0118] Depth feature maps are images obtained through camera intrinsic and extrinsic parameter calibration, coordinate transformation, and depth estimation. They reflect the spatial distance information of different regions in the image and can effectively distinguish between planar and three-dimensional structures.

[0119] For example, pixel coordinates can first be converted into physical coordinates with the image principal point as the origin based on camera intrinsic parameters, thus obtaining the true millimeter-level position on the imaging plane. Then, combined with camera extrinsic parameters (mount height, pitch angle, rotation matrix, translation vector), perspective projection and coordinate transformation are used to convert the physical coordinates of the image from the imaging plane coordinate system to a world coordinate system based on the ground / orbit plane, so that each pixel corresponds to a three-dimensional position in real space, thereby establishing a precise correspondence between image pixels and physical positions.

[0120] Furthermore, based on the pinhole imaging model, combined with the camera focal length, installation height, pitch angle, and image physical coordinates, monocular visual depth calculation is performed on each pixel to obtain the relative distance from that point to the camera optical center.

[0121] Then, based on the differences in the spatial structure of the carriage, the gap between the couplers, and the cargo, regional smoothing and concavity enhancement corrections are applied to the depth values. For example, the depth is kept uniform in the carriage area, the depth difference is enhanced in the gap between the couplers to highlight the concavity feature, and the natural undulations are preserved in the cargo area. Finally, a depth feature map is generated based on the normalized depth values ​​to distinguish between three types of spatial structures: flat, concave, and undulating.

[0122] S404 fuses the gradient map and the depth feature map to obtain a fused feature map.

[0123] A fused feature map is a comprehensive feature map obtained by fusing gradient maps and depth feature maps in a channel-weighted manner. It combines edge structure and spatial information, and has a stronger expressive power.

[0124] For example, the gradient map and the depth feature map can be concatenated along the channel dimension and then normalized and weighted to form a strong feature map that simultaneously contains structural edges and spatial depth.

[0125] S405, the fused feature map is input into the semantic segmentation model for semantic recognition to obtain the semantic mask features of the initial image data.

[0126] Semantic segmentation models refer to pixel-level classification models trained based on deep learning. They can take image features as input and output semantic masks for each category. They can use dedicated lightweight models to adapt to real-time detection.

[0127] For example, the fused features can be input into a semantic segmentation model, and the model outputs pixel-level classification results to form semantic mask features that separate the carriage, the gap between the coupler and the background.

[0128] In the above embodiments, color image features are extracted sequentially, gradient is calculated by grayscale to obtain a gradient map, coordinate transformation and depth calculation are performed to obtain a depth feature map, gradient and depth information are fused to obtain a fused feature map, and then the fused feature map is input into a semantic segmentation model to obtain semantic mask features. This can make full use of multi-dimensional information such as color, texture, edge, and spatial depth, and can still generate high-precision semantic masks in complex scenarios such as tunnel dim light, dust interference, and vehicle speed changes, reducing the misidentification and missed identification rate of the gap between the carriage and the coupling, and providing a stable and reliable foundation for subsequent boundary extraction and gap positioning.

[0129] In some alternative implementations, see [link to relevant documentation]. Figure 5 , Figure 5 A flowchart for obtaining dust detection results is provided, which includes the following steps:

[0130] S501, based on the cargo data in the segmented image data, determine the cargo area and the uncovered area of ​​the cargo surface without dust inhibitor.

[0131] Among them, the cargo area refers to the total pixel area occupied by the cargo region inside the carriage in the segmented image, which can be obtained by extracting the cargo region through semantic segmentation. The uncovered area refers to the total pixel area of ​​the cargo surface that is not covered by dust suppressant and is directly exposed, reflecting the degree of lack of dust suppressant spraying.

[0132] For example, cargo regions can be extracted from segmented image data, and the total pixel area of ​​the cargo can be counted; then, exposed cargo regions can be identified through color differences, and the pixel area of ​​uncovered inhibitors can be counted.

[0133] S502, the inhibitor test results are determined based on the cargo area and the uncovered area.

[0134] For example, the dust suppressant coverage can be judged as qualified or unqualified based on the ratio of the uncovered area to the cargo area, or based on the maximum single exposed area.

[0135] In some alternative implementations, if the ratio of the total uncovered area to the cargo area is greater than or equal to a first threshold, the inhibitor test result is determined to be unqualified, provided that the uncovered area is the total uncovered area.

[0136] Alternatively, if the uncovered area includes the independent uncovered areas of each dispersed region, and there exists an independent uncovered area that is greater than or equal to the second threshold, then the dust detection result is determined to be unqualified; wherein, the first threshold is greater than the second threshold.

[0137] The first threshold refers to the threshold for determining the ratio of the total uncovered area to the cargo area, used for overall coverage determination, and can be set to 10% for example. The second threshold refers to the threshold for determining the area of ​​a single scattered uncovered area, used for determining large local bare spots, and can be set to 5% for example. The first threshold is greater than the second threshold.

[0138] Total uncovered area refers to the sum of the areas of all exposed areas on the cargo surface that are not covered by dust suppressant. A single, independent uncovered area refers to a single, unconnected exposed cargo area identified through connectivity analysis, such as a large exposed area or a single spot where the dust suppressant was not applied.

[0139] For example, based on the cargo region in the segmented image data, the complete cargo region can be extracted first through a semantic segmentation model, and the total cargo area S_total can be obtained. Then, all exposed cargo regions without dust suppressant can be identified through color threshold and texture features, and the area of ​​each independent exposed region can be obtained, that is, the independent uncovered area of ​​each dispersed region. Finally, the total area of ​​all exposed regions can be obtained, that is, the total uncovered area.

[0140] Furthermore, if the ratio of the total uncovered area to the cargo area is greater than or equal to the first threshold, the inhibitor test result is determined to be unqualified. If at least one independent uncovered area is greater than or equal to the second threshold, the dust test result is determined to be unqualified.

[0141] In this way, by using a dual threshold of overall coverage and single-point exposed area for judgment, with the first threshold being greater than the second threshold, it can simultaneously take into account both the overall uniformity of dust suppressant coverage and the problem of severe local exposure, avoiding large-area missed spraying and local large spot missed spraying, making the detection and judgment more comprehensive, stricter, and more in line with environmental protection requirements.

[0142] S503 generates dust detection results based on the inhibitor detection results and the corresponding carriage markings in the segmented image data.

[0143] For example, the pass or fail conclusion can be bound to information such as carriage number, serial number, and collection time to generate structured test results for storage, uploading, and traceability.

[0144] For example, in a segmented image of a certain carriage, the calculated cargo area is 108,000 pixels, and the total area of ​​the exposed area on the cargo surface without dust suppressant is 9,500 pixels. Based on the area ratio, the suppressant test result is deemed qualified. Combined with the carriage identification "C64K-12345", a dust detection result containing the carriage number, the qualified conclusion, and the area ratio is generated.

[0145] In the above embodiments, the cargo area and uncovered area are first accurately calculated, and then the inhibitor coverage effect is quantitatively determined and bound to the carriage identification. This enables automated, standardized, and quantitative detection without the need for manual inspection and judgment, improving detection efficiency and consistency, while also generating traceable data.

[0146] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0147] Based on the same inventive concept, this application also provides a dust detection device for implementing the dust detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more dust detection device embodiments provided below can be found in the limitations of the dust detection method described above, and will not be repeated here.

[0148] In one exemplary embodiment, such as Figure 6 As shown, a dust detection device is provided, comprising:

[0149] The acquisition module 10 is used to acquire at least one frame of initial image data of the transport vehicle when the transport vehicle enters the detection area; wherein each frame of initial image data includes vehicle body data and cargo data.

[0150] The identification module 20 is used to identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments based on the vehicle body data.

[0151] The segmentation module 30 is used to segment each frame of initial image data acquired based on the identified car boundaries and the gap area between different cars to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car.

[0152] The detection module 40 is used to perform dust detection on the cargo data in each segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

[0153] In the above embodiments, by acquiring initial image data containing the vehicle body and cargo after the vehicle enters the detection area, accurately identifying the boundaries of the carriages and the gap between the couplers, and completing the image segmentation of a single carriage, the dust suppressant coverage detection is carried out on each carriage and the carriage identification is bound. This can upgrade the traditional coarse-grained detection of the entire train to fine-grained detection of a single carriage, quickly locate the position of the unqualified carriage, and realize the one-to-one correspondence between the detection results and the carriage information. This solves the problems of traditional detection being unable to accurately trace, difficult to locate, and inefficient.

[0154] In one embodiment, the identification module 20 is specifically used for:

[0155] For each initial image data, the initial image data is processed to obtain the semantic mask features corresponding to the initial image data; wherein, the semantic mask features include at least the mask features of the car body and the gap between the couplers; feature analysis is performed on each semantic mask feature to obtain the boundary of each car body of the transport vehicle and the gap area between different cars bodies.

[0156] In one embodiment, the identification module 20 is specifically used for:

[0157] The color image features of the initial image data are extracted; the color image features include color and texture information of the carriage and the gap between the coupler; the initial image data is converted to grayscale and gradient is calculated to obtain a gradient map containing gradient magnitude; the gradient map is transformed by coordinates and depth is calculated to obtain a depth feature map; the gradient map and the depth feature map are fused to obtain a fused feature map; the fused feature map is input into the semantic segmentation model for semantic recognition to obtain the semantic mask features of the initial image data.

[0158] In one embodiment, the identification module 20 is specifically used for:

[0159] Based on the mask features of the carriages in each semantic mask feature, the outermost contour of each carriage is determined; based on the outermost contour of each carriage, the boundary of each carriage of the transport vehicle is determined; connected component analysis is performed on the mask features of the hook gaps between different carriages in each semantic mask feature to obtain the hook gap regions between different carriages of the transport vehicle.

[0160] In one embodiment, the segmentation module 30 is specifically used for:

[0161] Using the gap between the car bodies as the car body boundary, each frame of the initial image data is segmented based on the identified car body boundary to obtain multiple segmented image data.

[0162] In one embodiment, the detection module 40 is specifically used for:

[0163] Based on the cargo data in the segmented image data, determine the cargo area and the uncovered area of ​​the cargo surface where dust inhibitors are not applied; based on the cargo area and uncovered area, determine the inhibitor detection results; based on the inhibitor detection results and the corresponding carriage markings in the segmented image data, generate the dust detection results.

[0164] In one embodiment, the detection module 40 is specifically used for:

[0165] If the total uncovered area is the total uncovered area, and the ratio of the total uncovered area to the cargo area is greater than or equal to a first threshold, then the inhibitor test result is determined to be unqualified; or, if the uncovered area includes the independent uncovered areas of each dispersed region, and there exists an independent uncovered area that is greater than or equal to a second threshold, then the dust test result is determined to be unqualified; wherein, the first threshold is greater than the second threshold.

[0166] Each module in the aforementioned dust detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0167] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores initial image data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a dust detection method.

[0168] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0169] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the dust detection method described in any of the above embodiments.

[0170] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the dust detection method described in any of the above embodiments.

[0171] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the dust detection method described in any of the above embodiments.

[0172] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0173] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0174] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0175] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dust detection method, characterized in that, The method includes: When a transport vehicle enters the detection area, at least one frame of initial image data of the transport vehicle is acquired; each frame of initial image data includes vehicle body data and cargo data. Based on the vehicle body data, identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments; Based on the identified car boundaries and the gap between different cars, each frame of initial image data is segmented to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car. For each segmented image data, dust detection is performed on the cargo data in the segmented image data to obtain dust detection results that include inhibitor detection results and carriage markings.

2. The method according to claim 1, characterized in that, The step of identifying the boundary of each compartment of the transport vehicle and the hook gap area between different compartments based on the vehicle body data includes: For each initial image data, the initial image data is processed to obtain the semantic mask features corresponding to the initial image data; wherein, the semantic mask features include at least the mask features of the carriage and the gap between the couplers; Feature analysis is performed on each of the semantic mask features to obtain the boundary of each compartment of the transport vehicle and the hook gap area between different compartments.

3. The method according to claim 2, characterized in that, The process of processing the initial image data to obtain the semantic mask features corresponding to the initial image data includes: Extract color image features from the initial image data; wherein, the color image features include color and texture information of the carriage and the gap between the couplers; The initial image data is processed to grayscale and gradient is calculated to obtain a gradient map containing gradient magnitudes. The gradient map is subjected to coordinate transformation and depth calculation to obtain a depth feature map; The gradient map and the depth feature map are fused to obtain a fused feature map; The fused feature map is input into a semantic segmentation model for semantic recognition to obtain the semantic mask features of the initial image data.

4. The method according to claim 2, characterized in that, The feature analysis of each of the semantic mask features to obtain the boundary of each compartment of the transport vehicle and the hook gap region between different compartments includes: Based on the mask features of the carriages in each of the semantic mask features, determine the outermost contour of each carriage; The boundary of each compartment of the transport vehicle is determined based on the outermost contour of each compartment. Connectivity analysis is performed on the mask features of the hook-and-gear gaps between different carriages in each of the semantic mask features to obtain the hook-and-gear gap regions between different carriages of the transport vehicle.

5. The method according to any one of claims 1-4, characterized in that, Based on the identified car boundaries and the gap regions between different cars, the initial image data of each frame is segmented to obtain multiple segmented image data, including: Using the gap between the car bodies as the car body boundary, each frame of the initial image data is segmented based on the identified car body boundary to obtain multiple segmented image data.

6. The method according to any one of claims 1-4, characterized in that, The step of performing dust detection on the cargo data in the segmented image data to obtain dust detection results including inhibitor detection results and carriage markings includes: Based on the cargo data in the segmented image data, determine the cargo area and the uncovered area of ​​the cargo surface without dust inhibitor; The inhibitor detection results are determined based on the cargo area and the uncovered area; Based on the inhibitor detection results and the carriage identification corresponding to the segmented image data, dust detection results are generated.

7. The method according to claim 6, characterized in that, The step of determining the inhibitor detection result based on the cargo area and the uncovered area includes: If the uncovered area is the total uncovered area, and the ratio of the total uncovered area to the cargo area is greater than or equal to a first threshold, then the inhibitor test result is determined to be unqualified; or, If the uncovered area includes the independent uncovered area of ​​each dispersed region, and there exists an independent uncovered area that is greater than or equal to the second threshold, then the dust detection result is determined to be unqualified; wherein, the first threshold is greater than the second threshold.

8. A dust detection device, characterized in that, The device includes: The acquisition module is used to acquire at least one frame of initial image data of the transport vehicle when the transport vehicle enters the detection area; each frame of initial image data includes vehicle body data and cargo data. The identification module is used to identify the boundary of each compartment of the transport vehicle and the hook gap area between different compartments based on the vehicle body data. The segmentation module is used to segment each frame of initial image data acquired based on the identified car boundaries and the gap area between different cars, to obtain multiple segmented image data; wherein, each segmented image data is the image data of one car, and the segmented image data includes the cargo data transported by the car. The detection module is used to perform dust detection on the cargo data in each segmented image data, and obtain dust detection results including inhibitor detection results and carriage markings.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images