A visual-based sintering grate bar end face quality detection method and device

Abstract

The application provides a sintering grate bar end face quality detection method and device based on vision, the method comprises the following steps: acquiring the end face image in the process of transporting the grate bar by a travelling crane as a first data image; performing edge detection on the preprocessed image, extracting the edge point information of the lower edge of the grate bar end face in the transverse direction, and fitting the lower edge line of the grate bar end face; detecting and positioning the edge feature marker on the grate bar travelling crane, and acquiring the circumscribed rectangle position information of the edge feature marker; when the first determination condition and the second determination condition are met at the same time, it is determined that the current first data image is the effective image of the grate bar end face with the maximum area, which is recorded as a second image; detecting the quality abnormal area on the grate bar end face, and outputting the category, position and size information of each abnormal area; comparing the size information of each abnormal area with the grading threshold value, grading the quality abnormality of the grate bar end face, and outputting the corresponding alarm information according to the abnormal grade.

Description

Technical Field

[0001] This application relates to the field of sintering monitoring technology, and in particular to a vision-based method and apparatus for detecting the end face quality of sintered grate bars. Background Technology

[0002] Sintering is one of the core processes in the iron and steel metallurgical production process. The sintering grate is a key component on the sintering trolley or overhead crane, used to support the sintering material and ensure the permeability of the sintering bed. Its working environment is characterized by high temperatures, high dust levels, mechanical vibration, and material erosion. Consequently, various quality abnormalities such as missing sections, loosening, burning, and deformation can gradually appear on the grate end faces. If these abnormalities are not promptly identified and addressed, they will directly affect the permeability and production stability of the sintering bed, reduce the quality of the finished sinter, and in severe cases, may even cause equipment failure, unplanned shutdowns, or even production safety accidents. Therefore, grate end face quality inspection is a crucial aspect of sintering production and maintenance.

[0003] Currently, the inspection of grate bar end face quality in sintering production sites mostly relies on regular manual inspections or manual checks after machine shutdown, with maintenance personnel visually assessing the grate bar's damage condition. This method has several drawbacks: First, manual inspection is labor-intensive, has a low frequency of inspections, and the results are heavily influenced by personnel experience and subjective judgment, leading to a high risk of missed or misjudged inspections. Second, inspections depend on machine shutdown or close-range manual observation, making continuous, full-coverage online inspection impossible during the grate bar's continuous operation, resulting in low inspection efficiency and potential operational safety hazards.

[0004] With the widespread application of machine vision technology in industrial inspection, there have been attempts to use visual imaging to inspect the condition of sintering bars. However, sintering sites generally suffer from problems such as drastic changes in lighting, significant dust interference, and severe shadow occlusion. Furthermore, as the sintering bar moves with the trolley, the imaging posture and projected area of ​​its end face in the image continuously change with its position. If defect identification is directly performed on the acquired full-frame image, it is highly susceptible to false positives or false negatives due to insufficient end face imaging area, misaligned edges, and viewing angle deviations. The accuracy and reliability of the detection cannot meet the actual application requirements of industrial sites. Summary of the Invention

[0005] In view of this, this application provides a vision-based method and device for detecting the end face quality of sintering grate bars, which can realize the refined identification of abnormal grate bar end face quality and differentiated operation and maintenance response, effectively ensuring the operational stability and production safety of sintering equipment while reducing operation and maintenance costs.

[0006] To achieve the above objectives, the first aspect of this application provides a vision-based method for inspecting the end face quality of sintered grate bars, comprising:

[0007] A camera positioned above the grate trolley with its lens facing the end face of the grate is used to acquire an end face image of the grate during transportation, which is then used as the first data image.

[0008] The first data image is preprocessed, and the image preprocessing includes at least color space conversion and equalization processing, and noise suppression processing;

[0009] Edge detection is performed on the preprocessed image to extract edge point information of the lower edge of the grate bar end face in the horizontal direction, and the lower edge line of the grate bar end face is fitted based on the edge point information.

[0010] The edge feature markers on the grate crane are detected and located using an object detection algorithm, and the position information of the circumscribed rectangle of the edge feature markers is obtained.

[0011] A reference anchor point is preset in the image coordinate system. The first distance between the edge feature marker and the reference anchor point is calculated, and the first distance not being greater than a first threshold is used as the first judgment condition. The second distance from the reference anchor point to the lower edge line of the grate end face is calculated, and the second distance not being greater than a second threshold is used as the second judgment condition. When the first judgment condition and the second judgment condition are satisfied at the same time, the current first data image is determined to be a valid image with the grate end face at its maximum area, and is recorded as the second image.

[0012] The second image is used to identify abnormal quality areas on the end face of the grate bar, and the category, location and size information of each abnormal area are output.

[0013] A preset threshold for abnormal size classification is set, and the size information of each abnormal area is compared with the classification threshold to classify the quality abnormalities of the grate end face, and corresponding alarm information is output according to the abnormality level.

[0014] As described above, the vision-based end-face quality inspection method for sintered grate bars provided in this application continuously acquires end-face images during the crane's operation using a camera. Image preprocessing is combined to eliminate environmental interference, and edge detection and target detection algorithms are used to extract the position information of the lower edge line of the grate bar end-face and the edge feature markers of the crane. Using a preset reference anchor point, a joint judgment is made based on dual spatial constraints: "distance between edge feature and anchor point" and "distance between lower edge line and anchor point." Only when both conditions are simultaneously met is the current frame considered a valid image where the grate bar end-face is at its maximum area. This method effectively avoids false detections and missed detections caused by changes in the end-face posture, viewing angle deviations, or the end-face not being directly facing the camera during the grate's movement with the crane. It ensures that subsequent defect detection is only performed on images where the end-face is fully exposed and in the optimal observation posture, significantly improving the consistency and accuracy of the detection results. Simultaneously, by automatically comparing the size of abnormal areas with grading thresholds, online automatic grading and alarming of end-face quality anomalies are achieved without manual intervention, improving the timeliness and safety of sintering equipment operation and maintenance.

[0015] As one possible implementation of the first aspect, the image preprocessing specifically includes:

[0016] The first data image is converted from the RGB color space to the HSV color space. In the HSV color space, the pixel brightness component is color balanced to reduce the effects of lighting changes and shadows.

[0017] Gaussian filtering is used to perform convolution processing on the color-equalized image to suppress noise and local bright spots in the image.

[0018] As shown above, by converting the image from RGB space to HSV space and performing color equalization on the luminance component, the hue and luminance information can be effectively separated, eliminating the uneven lighting caused by changes in fire, arc light, sunlight, and equipment shadows at the sintering site, making the image grayscale distribution more uniform. On this basis, Gaussian filtering is used to perform convolution smoothing on the image, which can effectively suppress the interference of particulate noise, camera electronic noise, and local bright spots on edge positioning in high dust environments.

[0019] As one possible implementation of the first aspect, Hough line detection is used to fit the edge point information into the lower edge line of the grate end face. The expression of the lower edge line of the grate end face in polar coordinate space is:

[0020] ;

[0021] in, The distance from the lower edge line to the origin of the image coordinate system is... Let be the angle between the normal vector of the lower edge line and the positive x-axis. , These are the coordinates of the edge points in the image.

[0022] Based on the above, the extracted lower edge point set of the grate bar end face is fitted into a complete lower edge line by using Hough line detection, and then plotted in polar coordinate space. and Two parameters are used to represent the straight line. The Hough transform has good noise resistance and robustness. Even if there is local occlusion, breakage or partial blurring at the lower edge, it can still fit the accurate position of the lower edge line based on the voting results of global edge points. The polar coordinate representation can uniformly describe straight lines in any direction and position, avoiding the numerical calculation problem of infinite slope of vertical straight lines in the Cartesian coordinate system, and laying a reliable foundation for the stable calculation of the distance from the reference anchor point to the edge line.

[0023] As one possible implementation of the first aspect, the position information of the bounding rectangle of the edge feature marker is represented as follows:

[0024] ;

[0025] in, , These are the image coordinates of the top-left corner of the circumscribed rectangle. , These are the width and height of the circumscribed rectangle, respectively.

[0026] As one possible implementation of the first aspect, the first distance is the Euclidean distance between the upper left corner of the bounding rectangle of the feature marker and the reference anchor point, calculated using the following formula:

[0027] ;

[0028] in, The first distance, , The image coordinates of the reference anchor point;

[0029] when When the first condition is met, where The first threshold is preset.

[0030] The second distance is the vertical distance from the reference anchor point to the lower edge line, and the calculation formula is:

[0031] ;

[0032] in, The second distance, , The image coordinates of the reference anchor point;

[0033] when When the second condition is met, where This is the preset second threshold.

[0034] As described above, the positional information of the edge feature markers is quantified using the coordinates of the upper left corner and the width and height of its circumscribed rectangle. The Euclidean distance between the upper left corner of the circumscribed rectangle and the reference anchor point is calculated as the first distance. The first judgment condition is quantified as the first distance not exceeding a first threshold. By substituting the coordinates of the reference anchor point into the polar coordinate equation of the lower edge line of the grate end face, the vertical distance from the anchor point to the lower edge line is calculated as the second distance. This second distance not exceeding a second threshold is used as the second judgment condition. This distance metric directly reflects the positional offset of the lower edge of the grate end face in the vertical direction of the image, accurately determining whether the end face has entered the optimal observation area in the center of the field of view. This calculation method involves only simple trigonometric function operations, resulting in high computational efficiency. Furthermore, it forms a dual complementary constraint of spatial position and orientation with the first judgment condition. The joint judgment of both ensures that the selected second image simultaneously satisfies the geometric condition of the maximum area of ​​the end face in both the horizontal and vertical directions.

[0035] As one possible implementation of the first aspect, the categories of the abnormal regions include missing grate bars, burned grate bars, and broken grate bars; the width of the rectangle enclosing the abnormal region... ,high Reference width of standard grate bar Reference height The comparison relationship distinguishes the abnormal categories:

[0036] like and If so, it is determined that the comb bar is missing;

[0037] like and If so, it is determined that the comb bars are burned.

[0038] like and If so, it is determined that the comb bars are damaged.

[0039] Based on the above, this classification strategy is based on the essential differences in physical deformation of different anomaly types. Missing leads to longitudinal stretching of the fracture surface (W becomes smaller and H becomes larger), burning leads to lateral flow of high-temperature melting (W becomes larger and H becomes smaller), and breakage leads to uniform shrinkage of the overall contour (both W and H become smaller). The anomaly category can be accurately identified by a simple width-height combination comparison. There is no need to train a complex classifier, the computational cost is extremely low, and the discrimination logic has clear physical interpretability.

[0040] As one possible implementation of the first aspect, the levels of the quality anomalies include minor anomalies, moderate anomalies, and severe anomalies;

[0041] When the size of the abnormal region is smaller than the first size threshold, it is determined to be a minor abnormality;

[0042] When the size of the abnormal region is greater than or equal to the first size threshold and less than the second size threshold, it is determined to be a moderate abnormality;

[0043] When the size of the abnormal region is greater than or equal to the second size threshold, or when the number of abnormal regions in the same second image reaches a preset number threshold, it is determined to be a serious abnormality, and alarm information including the abnormality category, abnormality level, abnormality location, acquisition time, and corresponding second image is output.

[0044] As described above, for detected abnormal areas, a three-level classification of minor, moderate, and severe anomalies is achieved by comparing the abnormal size with preset multi-level thresholds. This multi-dimensional classification mechanism effectively avoids the problem of underestimating a single size threshold when multiple small defects accumulate and cause serious safety risks, making the alarm strategy more comprehensive and rigorous. Different levels correspond to different maintenance response strategies; minor anomalies can be included in planned maintenance, while severe anomalies trigger immediate alarms, ensuring equipment safety while avoiding over-maintenance, which is conducive to optimizing spare parts replacement plans and reducing maintenance costs.

[0045] As one possible implementation of the first aspect, the edge feature marker on the grate trolley is a rigid structural component with stable position identification features in the edge of the trolley, a fixed sidewall, or a mounting bracket.

[0046] Based on the above, the edge feature markers on the grate trolley are defined as rigid structural components with stable position recognition features, such as the trolley edge, fixed sidewall, or mounting bracket. These components maintain a fixed relative position with the grate during trolley operation, and have regular shapes, clear edges, and are easy to be stably identified and located by target detection algorithms in the image.

[0047] As one possible implementation of the first aspect, it also includes:

[0048] If any frame of the first data image does not simultaneously satisfy the first condition and the second condition, the current frame image is discarded, and the camera is controlled to continue acquiring the next frame image until a second image that satisfies the first condition and the second condition is obtained.

[0049] Therefore, when the current frame image is determined not to meet the dual spatial constraints, the frame is automatically discarded and the camera continues to acquire the next frame image for judgment, until a second image that meets the conditions is successfully acquired. This frame filtering mechanism ensures that the defect detection algorithm only performs inference operations on images in the best observation pose, avoiding the waste of computing resources on invalid frames.

[0050] A second aspect of this application provides a vision-based sintered grate bar end face quality inspection device for implementing the aforementioned vision-based sintered grate bar end face quality inspection method, comprising:

[0051] The image acquisition module is used to acquire end face images of the grate during transportation by a camera set above the grate trolley with its lens facing the end face of the grate, and use these images as the first data image.

[0052] The preprocessing module is used to perform image preprocessing on the first data image, and the image preprocessing includes at least color space conversion and equalization processing, and noise suppression processing;

[0053] The edge feature extraction module is used to perform edge detection on the preprocessed image, extract the edge point information of the lower edge of the grate end face in the horizontal direction, and fit the lower edge line of the grate end face based on the edge point information.

[0054] The vehicle feature localization module is used to detect and locate edge feature markers on the grate vehicle using a target detection algorithm, and to obtain the position information of the bounding rectangle of the edge feature markers;

[0055] The maximum end-face frame filtering module is used to preset a reference anchor point in the image coordinate system, calculate the first distance between the edge feature marker and the reference anchor point, and use the first distance not being greater than a first threshold as a first judgment condition; calculate the second distance from the reference anchor point to the lower edge line of the grate end face, and use the second distance not being greater than a second threshold as a second judgment condition; when the first judgment condition and the second judgment condition are satisfied at the same time, the current first data image is determined to be a valid image with the grate end face at its maximum area, and is recorded as the second image;

[0056] The end face defect detection module is used to identify the second image using a target detection algorithm, detect quality abnormal areas on the end face of the grate bar, and output the category, location, and size information of each abnormal area;

[0057] The anomaly classification module is used to preset the anomaly size classification threshold, compare the size information of each anomaly area with the classification threshold, classify the quality anomalies of the grate end face into levels, and output corresponding alarm information according to the anomaly level. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating a vision-based method for inspecting the end face quality of sintered grate bars, as provided in an embodiment of this application.

[0059] Figure 2 This is a structural diagram of a vision-based sintered grate end face quality inspection device provided in an embodiment of this application. Detailed Implementation

[0060] To make this application easier to understand, specific embodiments are described below to further illustrate this application. The technical solutions provided by this application are further explained below with reference to the accompanying drawings and examples. It should be understood that the system architecture and business scenarios provided in the embodiments of this application are mainly for illustrating possible implementations of the technical solutions of this application and should not be construed as the sole limitation on the technical solutions of this application. Those skilled in the art will recognize that, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided by this application are equally applicable to similar technical problems.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. In case of any inconsistency, the meaning set forth in this specification or derived from the content described herein shall prevail. Furthermore, the terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit the scope of this application.

[0062] The solutions provided in this application will be described in detail below with reference to the accompanying drawings and embodiments.

[0063] This application proposes a vision-based method for inspecting the end face quality of sintered grate bars. This method can be applied to the grate bar transport operation in a sintering production line. The sintered grate bar array is installed on the trolley and moves continuously in a straight line along the track with the trolley. An industrial area array camera is installed directly above the grate bar trolley, with the camera lens facing the end face area of ​​the grate bar. The camera's field of view covers the entire movement stroke of the grate bar end face. The camera is connected to the field industrial control equipment via Ethernet, and the industrial control equipment executes image acquisition, processing, and detection algorithms to achieve online automated inspection of the grate bar end face quality.

[0064] Figure 1 The diagram shown is a flowchart of a vision-based method for inspecting the end face quality of sintered grate bars according to an embodiment of this application. (Refer to...) Figure 1 The method includes the following steps:

[0065] S110: A camera positioned above the grate trolley with its lens facing the end face of the grate is used to acquire an end face image of the grate during transportation, which is then used as the first data image.

[0066] In this embodiment, the camera's installation position should ensure that the grate end face can be completely displayed in the center of the camera's field of view when the grate trolley reaches the optimal observation position. During the continuous operation of the grate trolley, the camera continuously acquires real-time images of the grate end face at a preset frame rate, and each frame of the original acquired image is recorded as the first data image. Because the grate moves laterally with the trolley, the imaging position, projected area, and perspective orientation of the grate end face continuously change in different frames; when the grate end face is directly below the camera, the end face is parallel to the camera's imaging plane, resulting in the largest imaging area and the smallest perspective distortion, which is the optimal orientation for defect detection.

[0067] S120: Perform image preprocessing on the first data image, the image preprocessing including at least color space conversion and equalization processing, and noise suppression processing.

[0068] This step involves sequentially performing color space equalization and Gaussian filtering noise reduction on the first data image to eliminate interference from the on-site working conditions. Specifically:

[0069] First, the first data image in RGB format is converted to the HSV color space, and three independent channels are separated: hue (H), saturation (S), and lightness (V). Histogram equalization is performed only on the lightness (V) channel to stretch the grayscale dynamic range, thereby reducing the uneven grayscale distribution caused by ambient light fluctuations, equipment shadows, and localized high-temperature reflections. At the same time, the color and structural features of the grate end face are fully preserved to avoid color distortion.

[0070] Subsequently, a 5×5 Gaussian kernel is used to perform convolution filtering on the color-equalized image. The image is then smoothed by weighted averaging of neighboring pixels to suppress high-frequency interference caused by dust particles, camera thermal noise, and small material splashes, while preserving the sharpness of the edge face to the greatest extent, providing high-quality image input for subsequent edge detection and target detection.

[0071] S130: Perform edge detection on the preprocessed image, extract edge point information of the lower edge of the grate end face in the horizontal direction, and fit the lower edge line of the grate end face according to the edge point information.

[0072] For the preprocessed image, the Canny edge detection operator is used to extract the edge features of the whole image, and the horizontal edge point set corresponding to the lower edge of the grate end face is extracted by the horizontal gradient filtering; then the Hough line detection algorithm is used to fit the discrete edge point set with a straight line.

[0073] Hough line detection is based on the point-line duality principle, which identifies each edge point in the image space as ( Mapped to polar coordinates ( A sine curve in space, where This represents the perpendicular distance from the origin of the image coordinate system to the line. The angle between the normal vector of the line and the positive x-axis; the intersection of the sine curves corresponding to all edge points is the line parameter that is fitted by all edge points, and finally the polar coordinate expression of the lower edge line of the grate end face is obtained:

[0074]

[0075] Hough line detection can stably fit a complete lower edge line even if there is local dust obstruction, surface defects or noise interference at the lower edge of the end face, accurately representing the lateral attitude of the grate end face in the current frame.

[0076] S140: Detect and locate edge feature markers on the grate crane using a target detection algorithm, and obtain the position information of the circumscribed rectangle of the edge feature markers.

[0077] In this step, a pre-trained YOLOv11 object detection model is used to infer the preprocessed first data image, detecting and locating edge feature markers on the grate trolley. In this embodiment, the edge feature markers are fixed side rails on the side of the trolley. This component is a rigid structure, rigidly connected to the end face of the grate, moves synchronously, and has stable visual features, making it less susceptible to dust cover or material obstruction.

[0078] The target detection outputs the position information of the bounding rectangle of the edge feature marker:

[0079]

[0080] in, , These are the image coordinates of the top-left corner of the circumscribed rectangle. , These are the width and height of the circumscribed rectangle, respectively.

[0081] In some embodiments, edge feature markers may also be selected from other rigid structural components that are fixed in position and have stable features, such as vehicle mounting brackets or vehicle side edges.

[0082] S150: Set a first determination condition and a second determination condition. When the first determination condition and the second determination condition are met at the same time, determine that the current first data image is a valid image with the end face of the grate at its maximum area, and record it as the second image.

[0083] In this step, reference anchor points are pre-calibrated in the image coordinate system. The reference anchor point is determined through on-site calibration: when the end face of the grate bar is running directly below the camera, with the imaging area at its maximum and the orientation facing the camera, the standard position of the upper left corner of the outer rectangle of the edge feature marker and the standard position of the lower edge line of the end face are recorded, and the coordinate value of the reference anchor point is determined in this way.

[0084] Based on a reference anchor point, a dual-judgment condition is constructed to filter out valid images with the end face at the maximum area from a continuous image sequence:

[0085] First criterion: Calculate the Euclidean distance between the upper left corner of the circumscribed rectangle of the edge feature marker and the reference anchor point. As the first distance, the calculation formula is:

[0086]

[0087] Preset first threshold ,when When the crane has moved to the inspection station, the position condition is met, that is, the first judgment condition is met.

[0088] Second criterion: Calculate the perpendicular distance from the reference anchor point to the fitted lower edge line of the end face. As the second distance, the calculation formula is:

[0089]

[0090] Preset second threshold ,when When the grate bar end face is facing the camera, the posture condition is met, that is, the second judgment condition is met.

[0091] If and only if the first and second judgment conditions are met simultaneously, the current first data image is determined to be a valid image with the end face of the grate bar at its maximum area, and is recorded as the second image and sent to the subsequent defect detection steps; if either condition is not met, the current frame image is discarded, the next frame image is collected and the above judgment process is repeated until a second image that meets the requirements is selected.

[0092] In some embodiments, the first threshold Second threshold An adaptive adjustment mechanism can be used for adjustment; for example, when multiple consecutive frames do not meet the requirements... and The determination criteria indicate that the driving rhythm or camera installation status may have changed, at which point the first threshold can be automatically adjusted. Second threshold The threshold is increased by 1.1 to 1.2 times to adaptively adapt to changes in operating conditions. After successfully acquiring the second image, the threshold is restored to the original set value, thereby enhancing the adaptability of this method under different operating conditions.

[0093] S160: The second image is identified using a target detection algorithm to detect quality abnormalities on the end face of the grate bar, and the category, location, and size information of each abnormality area are output.

[0094] For the second image obtained from the screening, this step inputs the second image into the pre-trained end-face defect detection YOLOv11 model. This model is trained using end-face defect samples of the grate bars collected on-site and can identify three typical quality anomalies: missing grate bars, burned grate bars, and broken grate bars.

[0095] After inference, the model outputs the detection results of all abnormal regions in the second image. Each abnormal region includes: abnormality category, the position and size of its bounding rectangle, and detection confidence. The width of the bounding rectangle of each abnormal region is also output. ,high Reference width of standard grate bar Reference height The comparison relationship distinguishes the abnormal categories:

[0096] like and If so, it is determined that the comb bar is missing;

[0097] like and If so, it is determined that the comb bars are burned.

[0098] like and If so, it is determined that the comb bars are damaged.

[0099] S170: Preset abnormal size classification thresholds, compare the size information of each abnormal area with the classification thresholds, classify the quality abnormalities of the grate end face into levels, and output corresponding alarm information according to the abnormality level.

[0100] In this step, quality levels are determined by pre-setting two levels of anomaly size thresholds, namely the first size threshold and the second size threshold, combined with the size of the anomaly area and the number of anomalies:

[0101] When the equivalent size of the abnormal area is less than the first size threshold, it is judged as a minor abnormality, and only data is recorded and stored without triggering a real-time alarm.

[0102] When the equivalent size of the abnormal area is greater than or equal to the first size threshold and less than the second size threshold, it is determined to be a moderate abnormality, and the complete abnormal information is recorded and pushed to the operation and maintenance management system.

[0103] When the equivalent size of an abnormal area is greater than or equal to the second size threshold, or when the number of abnormal areas within the same end face reaches a preset number threshold, it is determined to be a serious abnormality, and an audible and visual alarm and a system alarm are immediately triggered. The output alarm information includes the abnormality category, abnormality level, abnormality location number on the end face, acquisition time, and corresponding second image screenshot, which are simultaneously pushed to the on-site central control system and maintenance personnel terminals to support differentiated maintenance and timely repair.

[0104] In summary, the vision-based end-face quality inspection method for sintered grate bars provided in this application embodiment can achieve fully automated online end-face quality inspection during the continuous operation of the sintered grate bar traveling with the trolley, without the need for machine shutdown or close-range manual inspection. This effectively solves the problems of high labor intensity, low inspection frequency, high risk of missed detection and misjudgment, and insufficient on-site safety of traditional manual inspection. Furthermore, through a dual-condition redundancy judgment mechanism combining reference anchor points, edge feature markers, and the lower edge line of the end face, it can accurately select standard posture images of the grate bar end face at the maximum imaging area, thus avoiding the problems caused by end-face viewing angle deviation and posture in conventional visual inspection from the data source. The problem of false positives and false negatives caused by tilting and insufficient imaging area is effectively addressed by a preprocessing scheme that combines HSV color space brightness equalization and Gaussian filtering. This scheme can effectively adapt to the complex working conditions of sintering site, such as severe light fluctuations, dust interference, and severe shadow occlusion, significantly improving the environmental robustness and accuracy of the detection results. In addition, the scheme can achieve refined identification and differentiated operation and maintenance response for grate end face quality anomalies through defect category quantification and multi-level size threshold anomaly classification alarm mechanism. Moreover, the entire scheme only requires a single camera for deployment, making installation and maintenance simple. While reducing operation and maintenance costs, it can effectively ensure the operational stability and production safety of sintering equipment.

[0105] like Figure 2 As shown, this application also provides a vision-based sintering grate end face quality inspection device. This device can be applied to the grate transport operation in a sintering production line to implement any step of the aforementioned vision-based sintering grate end face quality inspection method and its optional embodiments. (Refer to...) Figure 2 The device includes an image acquisition module 210, a preprocessing module 220, an edge feature extraction module 230, a vehicle feature localization module 240, a maximum end face frame filtering module 250, an end face defect detection module 260, and an anomaly classification module 270.

[0106] The image acquisition module 210 is used to acquire end-face images of the grate during transportation using a camera positioned above the grate trolley with its lens facing the end face of the grate, as the first data image; the preprocessing module 220 is used to perform image preprocessing on the first data image, the image preprocessing including at least color space conversion and equalization processing, and noise suppression processing; the edge feature extraction module 230 is used to perform edge detection on the preprocessed image, extract edge point information of the lower edge of the grate end face in the horizontal direction, and fit the lower edge line of the grate end face based on the edge point information; the trolley feature localization module 240 is used to detect and locate edge feature markers on the grate trolley using a target detection algorithm, and obtain the position information of the bounding rectangle of the edge feature markers; the maximum end-face frame filtering module 250 is used to preset a reference anchor point in the image coordinate system and calculate The first distance between the edge feature marker and the reference anchor point is used as a first judgment condition, where the first distance is not greater than a first threshold. The second distance from the reference anchor point to the lower edge line of the grate end face is calculated, and the second distance is not greater than a second threshold, which is used as a second judgment condition. When the first and second judgment conditions are met simultaneously, the current first data image is determined to be a valid image of the grate end face with the largest area, and is recorded as the second image. The end face defect detection module 260 is used to identify the second image using a target detection algorithm, detect quality abnormal areas on the grate end face, and output the category, location, and size information of each abnormal area. The abnormality classification module 270 is used to preset the abnormal size classification threshold, compare the size information of each abnormal area with the classification threshold, classify the quality abnormalities of the grate end face into levels, and output corresponding alarm information according to the abnormality level.

[0107] It should be understood that the apparatus or module in the embodiments of this application can be implemented by software, for example, by a computer program or instruction having the above-described functions. The corresponding computer program or instruction can be stored in the internal memory of the terminal, and the above functions can be implemented by the processor reading the corresponding computer program or instruction in the memory. Alternatively, the apparatus or module in the embodiments of this application can also be implemented by hardware. Or, the apparatus or module in the embodiments of this application can also be implemented by a combination of a processor and a software module.

[0108] It should be understood that the processing details of the apparatus or module in the embodiments of this application can be found by referring to... Figure 1 The descriptions of the embodiments and related extended embodiments shown will not be repeated in this application.

[0109] It should be noted that in the above description, the terms "first," "second," "third," etc., or similar terms such as module A, module B, and module C, used in the specification and claims are only used to distinguish similar objects and do not represent a specific order of objects. It is understood that, where permissible, a specific order or sequence can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein. In addition, the reference numerals indicating steps, such as S110, S120, etc., do not indicate that these steps must necessarily be performed. Where permissible, the order of preceding and following steps can be interchanged, or they can be performed simultaneously.

[0110] The term "comprising" as used in the specification and claims should not be construed as limiting itself to what follows; it does not exclude other elements or steps. Therefore, it should be interpreted as specifying the presence of the mentioned feature, integral, step, or component, but does not exclude the presence or addition of one or more other features, integrals, steps, or components, or groups thereof. Thus, the expression "device comprising means A and B" should not be limited to a device consisting solely of components A and B.

[0111] The terms "an embodiment" or "an embodiment" as used in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in at least one embodiment of this application. Therefore, the terms "in one embodiment" or "in an embodiment" appearing throughout this specification do not necessarily refer to the same embodiment, but may refer to the same embodiment. Furthermore, in one or more embodiments, the particular features, structures, or characteristics can be combined in any suitable manner, as will be apparent to those skilled in the art from this disclosure.

[0112] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of this application, all of which fall within the scope of protection of this application.

Claims

1. A vision-based method for inspecting the end face quality of sintered grate bars, characterized in that, include: A camera positioned above the grate trolley with its lens facing the end face of the grate is used to acquire an end face image of the grate during transportation, which is then used as the first data image. The first data image is preprocessed, and the image preprocessing includes at least color space conversion and equalization processing, and noise suppression processing; Edge detection is performed on the preprocessed image to extract edge point information of the lower edge of the grate bar end face in the horizontal direction, and the lower edge line of the grate bar end face is fitted based on the edge point information. The edge feature markers on the grate crane are detected and located using an object detection algorithm, and the position information of the circumscribed rectangle of the edge feature markers is obtained. In the image coordinate system, a reference anchor point is preset, and the first distance between the edge feature marker and the reference anchor point is calculated. The first distance is not greater than a first threshold as the first judgment condition. Calculate the second distance from the reference anchor point to the lower edge line of the end face of the grate bar, and use the second distance not being greater than the second threshold as the second judgment condition; When the first determination condition and the second determination condition are met simultaneously, the current first data image is determined to be a valid image with the end face of the grate at its maximum area, and is recorded as the second image; The second image is used to identify abnormal quality areas on the end face of the grate bar, and the category, location and size information of each abnormal area are output. A preset threshold for abnormal size classification is set, and the size information of each abnormal area is compared with the classification threshold to classify the quality abnormalities of the grate end face, and corresponding alarm information is output according to the abnormality level.

2. The method according to claim 1, characterized in that, The image preprocessing specifically includes: The first data image is converted from RGB color space to HSV color space. In the HSV color space, the pixel brightness component is color balanced to reduce the effects of lighting changes and shadows. Gaussian filtering is used to perform convolution processing on the color-equalized image to suppress noise and local bright spots in the image.

3. The method according to claim 1, characterized in that, The edge point information is fitted to the lower edge line of the grate bar end face using Hough line detection. The expression of the lower edge line of the grate bar end face in polar coordinate space is as follows: ； in, The distance from the lower edge line to the origin of the image coordinate system is... Let be the angle between the normal vector of the lower edge line and the positive x-axis. , These are the coordinates of the edge points in the image.

4. The method according to claim 3, characterized in that, The position information of the bounding rectangle of the edge feature marker is represented as follows: ； in, , These are the image coordinates of the top-left corner of the circumscribed rectangle. , These are the width and height of the circumscribed rectangle, respectively.

5. The method according to claim 4, characterized in that, The first distance is the Euclidean distance between the upper left corner of the circumscribed rectangle of the feature marker and the reference anchor point, calculated using the following formula: ； in, The first distance, , The image coordinates of the reference anchor point; when When the first condition is met, where The first threshold is preset; The second distance is the vertical distance from the reference anchor point to the lower edge line, and the calculation formula is: ； in, The second distance, , The image coordinates of the reference anchor point; when When the second condition is met, where This is the preset second threshold.

6. The method according to claim 1, characterized in that, The categories of abnormal areas include missing grate bars, burned grate bars, and broken grate bars; the width of the rectangle enclosing the abnormal area is... ,high Reference width of standard grate bar Reference height The comparison relationship distinguishes the abnormal categories: like and If so, it is determined that the comb bar is missing; like and If so, it is determined that the comb bars are burned. like and If so, it is determined that the comb bars are damaged.

7. The method according to claim 1, characterized in that, The quality anomalies are categorized into minor, moderate, and severe anomalies. When the size of the abnormal region is smaller than the first size threshold, it is determined to be a minor abnormality; When the size of the abnormal region is greater than or equal to the first size threshold and less than the second size threshold, it is determined to be a moderate abnormality; When the size of the abnormal region is greater than or equal to the second size threshold, or when the number of abnormal regions in the same second image reaches a preset number threshold, it is determined to be a serious abnormality, and alarm information including the abnormality category, abnormality level, abnormality location, acquisition time, and corresponding second image is output.

8. The method according to claim 1, characterized in that, The edge feature markers on the grate trolley are rigid structural components with stable position identification features in the edge of the trolley, fixed sidewall, or mounting bracket.

9. The method according to claim 1, characterized in that, Also includes: If any frame of the first data image does not simultaneously satisfy the first condition and the second condition, the current frame image is discarded, and the camera is controlled to continue acquiring the next frame image until a second image that satisfies the first condition and the second condition is obtained.

10. A vision-based sintered grate bar end face quality inspection device, used to implement the vision-based sintered grate bar end face quality inspection method according to any one of claims 1 to 9, characterized in that, include: The image acquisition module is used to acquire end face images of the grate during transportation by a camera set above the grate trolley with its lens facing the end face of the grate, and use these images as the first data image. The preprocessing module is used to perform image preprocessing on the first data image, and the image preprocessing includes at least color space conversion and equalization processing, and noise suppression processing; The edge feature extraction module is used to perform edge detection on the preprocessed image, extract the edge point information of the lower edge of the grate end face in the horizontal direction, and fit the lower edge line of the grate end face based on the edge point information. The vehicle feature localization module is used to detect and locate edge feature markers on the grate vehicle using a target detection algorithm, and to obtain the position information of the bounding rectangle of the edge feature markers; The maximum end-face frame filtering module is used to preset a reference anchor point in the image coordinate system, calculate the first distance between the edge feature marker and the reference anchor point, and use the first distance not being greater than a first threshold as the first judgment condition. Calculate the second distance from the reference anchor point to the lower edge line of the end face of the grate bar, and use the second distance not being greater than the second threshold as the second judgment condition; When the first determination condition and the second determination condition are met simultaneously, the current first data image is determined to be a valid image with the end face of the grate at its maximum area, and is recorded as the second image; The end face defect detection module is used to identify the second image using a target detection algorithm, detect quality abnormal areas on the end face of the grate bar, and output the category, location, and size information of each abnormal area; The anomaly classification module is used to preset the anomaly size classification threshold, compare the size information of each anomaly area with the classification threshold, classify the quality anomalies of the grate end face into levels, and output corresponding alarm information according to the anomaly level.

Images