A visual monitoring method and system for safe production of mine anchor bolts

Abstract

This invention relates to the field of image data processing technology, and more particularly to a visual monitoring method and system for the safe production of mine anchor bolts. The method includes the following steps: acquiring an image of the anchor bolt and performing multi-scale decomposition; extracting edge line segments within each scale layer and fitting them to obtain scale edge baselines; calculating the gradient consistency and edge sharpness indices of each scale baseline, and dynamically determining the fusion weights based on the scale size; performing weighted fusion of all scale baselines to obtain a global edge baseline; calculating the straightness of the global edge baseline using a straightness model that comprehensively evaluates positional deviation, parallelism, and spacing, and determining the safety risk of the anchor bolt accordingly. This invention, through weighted fusion, can suppress interference from local noise and false edges, improving the accuracy and robustness of anchor bolt straightness detection.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology, and in particular to a visual monitoring method and system for safe production of mine anchor bolts. Background Technology

[0002] Mining anchor bolts are core load-bearing components in mine shaft and tunnel support engineering. They are mainly used to reinforce surrounding rock and prevent rock mass collapse. Their structural stability directly determines the safety factor and engineering durability of mining operations. Therefore, there are strict quality standards for the straightness, strength and other physical properties of the anchor bolt body.

[0003] In the production process of mining anchor bolts, the bolt body is prone to quality defects after the straightening process due to multiple factors. On the one hand, the straightening equipment itself has precision deviations, and when the equipment has been running for a long time, parts wear out and parameters drift, making it difficult to guarantee straightening accuracy. On the other hand, during manual operation, deviations can easily occur in the control of straightening force and the adjustment of the bolt clamping angle. The combination of these two factors often leads to over-correction or localized stress concentration deformation of the bolt body. These latent defects will quickly become apparent when subjected to force in subsequent processes, causing the bolt body to spring back and bend, forming explicit quality problems.

[0004] Currently, the straightness detection of anchor bolts is typically achieved using the LSD (Long Segment Deposition) straightness detection algorithm. However, this technology has significant limitations in practical applications and struggles to meet high-precision detection requirements. Firstly, due to the limited installation space in industrial inspection scenarios, the shooting distance between the camera and the anchor bolt under test is usually quite far, resulting in relatively low resolution images of the bolt. The LSD algorithm easily extracts broken or discontinuous edge segments, failing to fully capture the overall trend of the bolt's axis and thus missing some bending defects. Secondly, the LSD algorithm processes images at a single scale, making it difficult to simultaneously capture the large-scale overall outline of the bolt and small-scale local details. While macroscopic analysis can grasp the overall direction of the anchor bolt, it easily overlooks minor local bends. Conversely, microscopic analysis, while capturing local details, is highly susceptible to interference from surface textures, burrs, and other noise, losing its ability to judge the overall trend, leading to unstable detection results and poor adaptability. Summary of the Invention

[0005] To address the limitations of the LSD straight line detection algorithm, such as edge line breakage due to long shooting distance and the inability of a single scale to simultaneously capture large-scale contours and small-scale details, which leads to unstable anchor bolt straightness detection results and difficulty in meeting high-precision requirements, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a visual monitoring method for safe production of mine anchor bolts, the method comprising the steps of:

[0007] An image of the anchor rod to be tested is acquired, and edge detection is performed on the image to obtain the edge image of the anchor rod. The edge image is decomposed into multiple scales to obtain scale images at several scales. For each scale image, line detection is performed to extract multiple line segments, and fitting is performed based on the multiple line segments to obtain the edge baseline at the corresponding scale. The edge baselines at all scales are weighted and fused to obtain the global edge baseline. The weight of the weighted fusion is positively correlated with the product of the gradient consistency and edge sharpness of the pixels on the edge baseline at the corresponding scale, and is also positively correlated with the scale size. The straightness of the global edge baseline is calculated. The straightness is positively correlated with the average distance from the global edge baseline to the edge of a preset qualified anchor rod, the slope difference of the global edge baseline, and the average spacing. The safety status of the anchor rod to be tested is determined based on the straightness of the global edge baseline.

[0008] This invention fundamentally solves the problem of unstable results caused by local noise and defects in single-scale detection by introducing a multi-scale decomposition and weighted fusion strategy. Existing technologies often treat multi-scale information equally or use simple weighting when fusing multi-scale information, failing to distinguish information quality. This invention, however, uses a weighting method based on gradient consistency, edge sharpness, and scale size to quantitatively assess the reliability of detection results at each scale layer, assigning greater weight to high-quality information and effectively suppressing noise and false edges. Simultaneously, its straightness calculation model integrates information from three dimensions: positional deviation, parallelism, and spacing, providing a more comprehensive and accurate assessment than a single indicator. Ultimately, this achieves high-precision and robust detection of anchor bolt straightness, significantly improving the reliability of safety monitoring on automated production lines.

[0009] Preferably, the step of fitting based on the plurality of straight line segments to obtain the edge baseline at the corresponding scale includes: calculating the probability that each straight line segment belongs to a valid straight line segment based on the angle difference between each straight line segment and the main axis direction of the anchor rod to be tested and the ratio of its length to the average length of all straight line segments at the current scale; determining the straight line segments with the probability greater than a preset probability threshold as valid straight line segments, and fitting based on the valid straight line segments to obtain the edge baseline.

[0010] This invention adds a probability-based effective line segment screening step, which can effectively remove obviously irrelevant noise and abnormal line segments based on the angle and relative length of the line segment to the principal axis, ensuring the quality of the data used for subsequent fitting, thereby improving the accuracy of edge baseline fitting at various scales.

[0011] Preferably, the probability that the line segment belongs to a valid line segment is... Satisfying the relation:

[0012] ;

[0013] in, It is the first at the current scale The angle between the straight line segment and the main axis direction of the anchor bolt; It is the direction of the anchor bolt's main axis; It is the first at the current scale The length of a straight line segment; It is the average length of all straight line segments at the current scale; It is a natural exponential function; It is a standard normalized function; It is the absolute value symbol.

[0014] This invention provides a specific and repeatable mathematical basis for selecting effective straight line segments by penalizing angular deviations through a negative exponential function and combining it with a normalized length ratio, making the selection process more accurate and stable.

[0015] Preferably, the acquisition of gradient consistency includes: calculating the sum of squares of the differences between the local gradients of all pixels on the edge baseline at the same scale and the average gradient of the edge baseline, and recording the result of negatively normalizing the sum of squares as gradient consistency.

[0016] Preferably, the edge sharpness satisfies the following relationship:

[0017] ;

[0018] in, It is the first The first layer at the layer scale Edge sharpness of the baseline; It is the first The first layer at the layer scale The number of valid straight line segments of the edge baseline; This is the preset maximum number of valid straight line segments; It is the first Layer scale The average gradient magnitude of all pixels along the edge baseline; It is the average gradient magnitude of the preset pixels.

[0019] Preferably, the step of weighted fusion of edge baselines at all scales to obtain a global edge baseline includes: each edge baseline at each scale includes an upper edge baseline and a lower edge baseline; for pixels on the upper edge baselines at all scales, a global upper edge baseline is obtained by least squares fitting based on the weights of the weighted fusion; for pixels on the lower edge baselines at all scales, a global lower edge baseline is obtained by least squares fitting based on the weights of the weighted fusion; the global upper edge baseline and the global lower edge baseline together constitute the global edge baseline.

[0020] This invention clarifies the specific implementation method of weighted fusion, namely, performing independent weighted least squares fitting on the upper and lower edge baselines separately. Compared with the possible methods of coupling the two edges or fitting a center line, this independent processing method simplifies the calculation model and can better adapt to the complex situation of inconsistent quality between the upper and lower edges, thus improving the robustness and accuracy of global edge baseline generation.

[0021] Preferably, the straightness of the global edge reference line Satisfying the relation:

[0022] ;

[0023] in, , These are the average distances from all pixels on the global top and bottom edge baselines to the edge of the preset qualified anchor bar; , These are the slopes of the global upper and lower edge baselines, respectively; It is the average distance between the global upper and lower edge baselines; It is the average distance between the upper and lower edge reference lines of the pre-set qualified anchor rod; It is a standard normalized function; It is the absolute value symbol.

[0024] This invention provides a clear quantitative model for the final straightness assessment. Compared to traditional methods that rely on a single indicator, this formula integrates three key geometric features: the absolute positional deviation of the edge, the parallelism of the two edges, and the average diameter of the anchor bolt. By normalizing and summing these multi-dimensional information, a comprehensive evaluation score is formed, making the final assessment of the anchor bolt's safety status more reliable and complete.

[0025] Preferably, the multi-scale decomposition of the edge image is achieved using the Laplacian pyramid decomposition algorithm.

[0026] Preferably, the line detection is implemented using the LSD line detection algorithm.

[0027] In a second aspect, the present invention provides a visual monitoring system for safe production of mining anchor bolts. The visual monitoring system for safe production of mining anchor bolts includes a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the visual monitoring method for safe production of mining anchor bolts according to the first aspect of the present invention is implemented.

[0028] By adopting the above technical solution, a computer program is generated from the first aspect of the present invention for a visual monitoring method for safe production of mine anchor bolts, and stored in a memory so that it can be loaded and executed by a processor. A terminal device is then made based on the memory and the processor for convenient use.

[0029] The beneficial effects of this invention are as follows: First, it overcomes the limitations of single-scale detection through multi-scale decomposition. Then, it proposes a weighted fusion strategy based on gradient consistency and edge sharpness indices to quantitatively evaluate the reliability of information at each scale level, thereby effectively suppressing noise and false edge interference during fusion. Finally, in the linear quantification model, it integrates three key geometric features: the absolute positional deviation of the edge, the parallelism of the two edges, and the average diameter of the anchor bolt. By normalizing and summing these multi-dimensional information, a comprehensive evaluation score is formed, making the final judgment on the safety status of the anchor bolt more reliable and comprehensive. This achieves high-precision and high-robustness judgment of the anchor bolt's safety status. Attached Figure Description

[0030] Figure 1 A flowchart of a visual monitoring method for safe production of mine anchor bolts provided in an embodiment of the present invention;

[0031] Figure 2 This is a 0th-level scale image of the Laplacian pyramid scale decomposition provided in an embodiment of the present invention;

[0032] Figure 3 This is the first-level scale image after Laplacian pyramid scale decomposition provided in an embodiment of the present invention;

[0033] Figure 4 This is a second-level scale image provided by an embodiment of the present invention after Laplacian pyramid scale decomposition;

[0034] Figure 5 This is a third-level scale image provided by an embodiment of the present invention after Laplacian pyramid scale decomposition;

[0035] Figure 6 This is a structural block diagram of a visual monitoring system for safe production of mine anchor bolts, provided as an embodiment of the present invention. Detailed Implementation

[0036] The first aspect of this invention provides a visual monitoring method for the safe production of mine anchor bolts, such as... Figure 1 As shown, the method includes steps S100-S600:

[0037] Step S100: Obtain the image of the anchor rod to be tested, and perform edge detection on the image of the rod to obtain the edge image of the anchor rod to be tested.

[0038] It should be noted that acquiring the image of the anchor bolt under test is the data foundation of the entire visualization monitoring process, and its quality directly affects the accuracy of all subsequent analyses. Edge detection of the bolt image is necessary to extract the contour information most relevant to the anchor bolt's shape from the complex background and surface texture, forming concise and crucial geometric features. This is a prerequisite for subsequent linearity quantification calculations.

[0039] Specifically, one or more industrial high-speed cameras can be vertically positioned directly above the conveyor belt after the straightening process in the mining anchor bolt production line and before the next process. These cameras acquire images of the anchor bolts being tested immediately after straightening. After obtaining the raw bolt images, they are preprocessed. Preferably, the acquired color or grayscale images are uniformly converted into single-channel grayscale images to reduce computational complexity.

[0040] Subsequently, the Canny edge detection algorithm was used to process the grayscale image. The Canny algorithm effectively extracts the edge pixels of the anchor rod through a series of operations, including Gaussian filtering to smooth the image, calculating gradient magnitude and direction, non-maximum suppression to thin the edges, and double thresholding to connect the edges. To further improve the continuity and integrity of the edges, a morphological closing operation (dilation followed by erosion) can be performed on the Canny detection results to fill in minute breaks in the edges and smooth the contours.

[0041] Thus, an edge image that clearly and completely represents the geometric contour of the anchor rod under test has been obtained.

[0042] Step S200: Perform multi-scale decomposition on the edge image to obtain scale images at several scales.

[0043] It should be noted that single-scale image analysis has inherent limitations: macroscopic analysis can accurately capture the overall bending trend of the anchor bolt, but it ignores minute local deformations; while microscopic analysis, although sensitive to local details, is easily affected by pixel-level noise, making it difficult to form a stable judgment on the global morphology. Therefore, this invention adopts a multi-scale decomposition strategy, decomposing the edge image to scales with different spatial frequencies, in order to capture the morphological features of the anchor bolt at different levels.

[0044] This embodiment preferably employs the Laplacian pyramid decomposition algorithm. This algorithm is an efficient, overcomplete image decomposition method. Its core idea is to construct a Gaussian pyramid by Gaussian smoothing and downsampling the original image, then upsampling a layer of the Gaussian pyramid and performing a difference operation with the previous layer. The resulting difference image is one layer of the Laplacian pyramid. Each layer of the Laplacian pyramid represents information from the original image in a specific frequency band. Its advantages lie in its ability to effectively separate the structural and detailed information of the image, and the computational efficiency of the decomposition and reconstruction process. This algorithm was chosen because it can effectively decouple the macroscopic contours and local edge details of the anchor rod into images of different scales.

[0045] Specifically, firstly, the edge image obtained in step S100 is used as the input to the 0th layer of the Laplacian pyramid. After iterative Gaussian filtering and downsampling operations, a layer containing... The Gauss Pyramid of 100 layers ,in, It is the Pyramid of Gauss. The layer, It is the Pyramid of Gauss. The Layer. Secondly, through the... Perform upsampling and Difference, generating the first of Laplace's pyramids Scale image corresponding to the layer .

[0046] Finally, repeating this process will yield a set of scaled images. ,in, It contains the finest edge details, and This reflects the most macroscopic outline of the anchor bolt. In this embodiment, the number of decomposition layers... The optimal value can be 3 to 5. Furthermore, the Laplace pyramid decomposition algorithm is existing technology, and its detailed process will not be elaborated upon here.

[0047] like Figures 3-5 The image shown is an image of each scale after Laplacian pyramid scale decomposition when the number of decomposition levels is set to 3. Figure 2 It is the edge image obtained in step S100, corresponding to the 0th layer of the Laplacian pyramid; Figure 3 , Figure 4 , Figure 5 These correspond to the scale images of layer 1, layer 2, and layer 3, respectively. Observe. Figures 3-5 It can be seen that, Figure 3 The edge details of the middle anchor are still relatively clear. As the scale increases, the image gradually focuses on macroscopic features, reaching the third layer. Figure 5 At that time, it only reflected the most macroscopic outline of the anchor bolt.

[0048] Thus, the multi-scale decomposition of the edge image was completed, and several scale images capable of representing the anchor bolt contour features at different granularities were obtained.

[0049] Step S300: For each scale image, perform line detection to extract multiple line segments, and fit based on the multiple line segments to obtain the edge baseline at the corresponding scale.

[0050] It should be noted that after multi-scale decomposition, the anchor edge contour in images at each scale may still consist of discontinuous line segments. To accurately describe the overall trend of the anchor edge at that scale, these discrete line segments need to be integrated. Line detection algorithms can effectively extract all line segment elements from the scale image, but they include pseudo-line segments caused by noise or anchor surface texture. Therefore, it is necessary to first screen out the effective line segments that truly reflect the anchor edge, and then perform least-squares fitting based on these effective line segments to obtain an edge baseline that represents the overall trend of the edge at that scale.

[0051] Specifically, for each scale image obtained in step S200 The specific operations include: steps S310-S330:

[0052] Step S310: Perform line detection on each scale image.

[0053] Preferably, the LSD (Line Segment Detector) algorithm is used to detect lines in each scale image. The LSD algorithm requires no manual parameter tuning and can quickly and accurately detect local line segments in an image, providing information such as endpoint coordinates and length, making it highly suitable for industrial automation scenarios. After executing the LSD algorithm, an initial set of line segments at that scale is obtained.

[0054] Step S320: Filter the valid line segments in the initial set of line segments.

[0055] It should be noted that the initial set of line segments may contain invalid segments caused by noise or artifacts. To obtain more reliable fitting results, these line segments need to be screened. The screening criteria are: line segments belonging to the edge of the real anchor bolt should have a direction approximately parallel to the principal axis of the anchor bolt being measured, and their length should be significant.

[0056] Based on the above logic, the probability of a valid line segment satisfies the following relationship:

[0057] ;

[0058] in, It is the first at the current scale The probability that a line segment is a valid line segment; It is the first at the current scale The angle between the straight line segment and the main axis direction of the anchor bolt; This refers to the direction of the anchor bolt's main axis. If the anchor bolt is placed horizontally in the image, then... ; It is the first at the current scale The length of a straight line segment; It is the average length of all straight line segments at the current scale; It is a natural exponential function; It is a standard normalized function; It is the absolute value symbol.

[0059] In this relation, It is the first at the current scale The angle between the straight line segment and the principal axis of the anchor rod; the smaller this value, the smaller the angular deviation between the straight line segment and the principal axis of the anchor rod. The closer it is to 1, the higher its contribution to the probability. It is the first at the current scale The ratio of the length of this straight line segment to the average length of all straight line segments at the current scale. Map this ratio to Interval. The longer the average length, The larger the value of , the higher its contribution to the probability. This relationship reflects that the more parallel a straight line segment is to the main direction and the longer it is, the higher the probability that it represents a true edge.

[0060] It should be noted that to filter valid line segments, after calculating the probability of each line segment, a preset probability threshold needs to be set. All line segments with a probability greater than this threshold are then considered valid. Regarding the setting of the preset probability threshold, setting it too high... At times, stringent screening conditions, while retaining high-purity, noise-free straight line segments, easily exclude edge segments corresponding to genuine slight bends or local defects in the anchor rod, resulting in insufficient effective straight line segments. This leads to overly smooth subsequent fitting results and potential missed detections. Conversely, if the threshold is too low (<0.5), the screening conditions are too lenient, easily incorporating background noise, rust marks on the rod, and other pseudo-straight line segments, interfering with the fitting process and causing deviations or abnormal bends in the edge baseline, potentially resulting in false alarms. This embodiment sets the probability threshold to 0.7, which implementers can adjust according to their needs.

[0061] Step S330: Fit the selected valid line segments to obtain the edge baseline.

[0062] The midpoint coordinates of all valid straight line segments are extracted. Taking a horizontally placed anchor as an example, the anchor typically has upper and lower edges, so the ordinates of these midpoints naturally form two clusters. The K-means clustering algorithm is used to divide these midpoints into an upper edge point set and a lower edge point set. Then, for the endpoints of all valid straight line segments belonging to the upper and lower edge point sets, the least squares method is used for linear fitting, ultimately obtaining the upper edge baseline equation and the lower edge baseline equation at this scale. These two baselines together constitute the edge baseline at the corresponding scale.

[0063] Thus, for each scale image, the upper and lower edge baselines representing the anchor bolt profile at that scale can be obtained.

[0064] Step S400: Weighted fusion of edge baselines at all scales to obtain global edge baselines.

[0065] It should be noted that the edge baselines at different scales have different emphases. The baseline at the macro scale reflects the global trend, while the baseline at the micro scale supplements local details. In order to obtain a final outline that best represents the true shape of the anchor bolt and is both macroscopically accurate and rich in detail, information from all scales must be integrated. Simple average integration will treat all information equally, while weighted integration can assign different importance according to the reliability of information at each scale, thereby obtaining a better global result.

[0066] Specifically, the fusion process includes two stages: weight calculation and weighted fitting.

[0067] Regarding the weight calculation, it should be noted that the weights of the weighted fusion are positively correlated with the product of the gradient consistency and edge sharpness of the pixels on the edge baseline at the corresponding scale, and are also positively correlated with the scale size.

[0068] First, regarding gradient consistency, it measures the smoothness of a baseline. An ideal straight line should have consistent local gradients across all pixels. Therefore, obtaining gradient consistency involves calculating the sum of squares of the differences between the local gradients of all pixels on the edge baseline at the same scale and the average gradient of the edge baseline. The result of negatively normalizing this sum of squares is denoted as gradient consistency; the smaller this value, the higher the gradient consistency.

[0069] Secondly, regarding edge sharpness, this invention considers that for an ideal anchor edge, the LSD algorithm should detect a small number of relatively long straight line segments. However, in actual working conditions, due to factors such as uneven lighting and minor imperfections on the anchor surface, a continuous edge may be incorrectly segmented into a large number of fragmented short straight line segments by the LSD algorithm. Although these short straight line segments may themselves meet the selection criteria for angle and length, meaning they are all valid straight line segments, an excessive number of valid straight line segments reflects poor edge continuity and a decrease in quality.

[0070] Therefore, this invention incorporates consideration of the number of effective straight line segments when calculating edge sharpness. The edge sharpness constructed by this invention satisfies the following relationship:

[0071] ;

[0072] in, It is the first The first layer at the layer scale Edge sharpness of the baseline; It is the first The first layer at the layer scale The number of valid straight line segments of the edge baseline; This is the preset maximum number of valid straight line segments; It is the first Layer scale The average gradient magnitude of all pixels along the edge baseline; It is the average gradient magnitude of a preset pixel; in this embodiment This includes the upper edge baseline and the lower edge baseline.

[0073] In this relationship, edge sharpness is related to... The results are directly proportional, which means that, given the same edge strength, the number of effective straight line segments constituting the edge baseline will increase. The less, The higher the value, the higher the calculated edge sharpness, and vice versa. In this way, the present invention can more accurately assess the overall quality of the edge baseline and effectively distinguish truly continuous, high-quality edge baselines.

[0074] It should be noted that the maximum number of valid straight line segments is preset. Its setting aims to define the boundary between continuity and fragmentation. The method for obtaining this value includes: acquiring a large number, for example, hundreds, of qualified anchor bolt sample images that meet quality requirements; performing steps S100-S300 on these sample images under the same lighting and camera settings as the actual inspection; and statistically determining how many effective straight line segments each qualified edge consists of at each scale. Composition. Typically, a high-quality continuous edge will be detected by LSD as 1 to 3 longer line segments. To tolerate a certain degree of normal fluctuation, This should be set to an integer slightly larger than the maximum number of line segments observed in these qualified samples. For example, if statistics show that the qualified samples... In most cases, the value does not exceed 4. It can be preferably set to 5 or 6. This setting ensures that... Edges composed of 5 line segments are given a high continuity score, while broken edges composed of more than 5 line segments are significantly penalized.

[0075] For the preset maximum average gradient magnitude Its function is to provide a normalized upper bound for edge strength. The method for obtaining this value is similar to... Similarly: Under the same qualified sample set and imaging conditions, calculate the average gradient magnitude on all qualified edge baselines. . This should be set to a value slightly higher than the maximum average gradient magnitude observable in these qualified samples. This value is closely related to factors such as the image's bit depth (e.g., 8-bit grayscale), illumination intensity, and camera gain. For example, in a typical 8-bit grayscale... In the imaging system, experimental measurements show that the average gradient amplitude of a clear, sharp anchor bolt edge is typically around 180. Considering potential strong reflections, etc., It can be preferably set to 250 to ensure that in most cases... The ratio falls on Within the interval, thus achieving effective normalization.

[0076] Then, the final weights are calculated by combining the product of the gradient consistency and edge sharpness of pixels on the edge baseline with the scale. Weights of weighted fusion at layer scale Satisfying the relation:

[0077] ;

[0078] in, It is the first The first layer at the layer scale Gradient consistency of the edge baseline; It is the first The first layer at the layer scale Edge sharpness of the baseline; It represents the total number of decomposition levels in the scale decomposition. It is the scale level of the currently calculated image. It is the first In this embodiment, the number of edge baselines at the layer scale is... ; It is the standard normalization function.

[0079] In this relation, As a scale priority factor, ensure the scale hierarchy The larger the value, the higher the base weight. Ultimately, the more reliable and larger the scale of the edge baseline, the higher the weight will be assigned.

[0080] At this point, we have obtained the weights needed for weight fusion at all scales. Next, we will perform the fitting process.

[0081] Specifically, a weighted least squares method is used for fitting, with the fusion process performed separately for the top and bottom edges. Taking the fitting of the global top edge baseline as an example, all pixels on the top edge baseline at all scales are assigned weights corresponding to their respective scales. Then, the weighted least squares method is used to fit these weighted point sets to obtain the slope of the global top edge baseline. and intercept .

[0082] Similarly, a weighted fit is performed on all lower edge baselines to obtain the global lower edge baseline slope. and intercept These two lines together form the final global edge baseline.

[0083] Thus, through multi-scale information weighted fusion, a global edge baseline that can comprehensively and accurately reflect the overall outline of the anchor rod under test was obtained.

[0084] Step S500: Calculate the straightness of the global edge baseline.

[0085] It should be noted that although the global edge baseline is a precise mathematical description of the anchor bolt profile, it is still two straight line equations. To achieve automated judgment, this geometric description needs to be transformed into a single, quantitative indicator that can intuitively reflect its straightness, namely, straightness.

[0086] Specifically, for effective comparison, this invention first needs to establish an ideal mathematical model of a pre-defined qualified anchor rod. It is understood that the pre-defined qualified anchor rod here does not refer to a physical entity, but rather a set of benchmark parameters pre-defined and stored within the system before actual testing. This ideal model represents a perfect anchor rod that is absolutely straight, has a constant diameter, and conforms to design specifications in the current imaging coordinate system. The establishment of this model is usually completed during system initialization or when the specifications of the anchor rod to be tested are changed on the production line. Its specific parameters include: Pre-defined qualified anchor rod edge lines: In the image coordinate system, two completely parallel and fixed-position straight lines are defined as the ideal upper and lower edges. For example, in a system where anchor rods are placed horizontally, these two lines can be defined as... and ,in and This is a fixed vertical axis constant. Preset qualified anchor diameter: This parameter is the distance between the upper and lower edges of a preset qualified anchor. It originates from the design specifications of the anchor to be tested, for example, a physical diameter of 20mm. During the system calibration phase, it is necessary to establish the conversion relationship between physical dimensions and image dimensions by photographing a calibration object with known physical dimensions. This refers to the pixel value corresponding to the design diameter, obtained based on this conversion relationship, and its value is greater than zero.

[0087] After establishing the aforementioned ideal model, the straightness proposed in this invention considers three aspects of deviation: overall curvature, parallelism, and width uniformity. Its construction logic is as follows: for an ideal anchor bolt, its global edge reference line should coincide with the edge of a pre-defined qualified anchor bolt, and the two lines should be parallel to each other with a spacing equal to the standard diameter. Based on this, the straightness of the global edge reference line... Satisfying the relation:

[0088] ;

[0089] in, , These are the average distances from all pixels on the global top and bottom edge baselines to the edge of the preset qualified anchor bar; , These are the slopes of the global upper and lower edge baselines, respectively; It is the average distance between the global upper and lower edge baselines; It is the average distance between the upper and lower edge reference lines of the pre-defined qualified anchor rod, corresponding to the pre-defined qualified anchor rod diameter; It is a standard normalized function; It is the absolute value symbol.

[0090] In this relation, the first part The macroscopic bending degree of the anchor bolt was quantified in Part Two. This quantifies the morphological regularity of the anchor bolt itself, namely its parallelism and symmetry. Among these, It is the absolute value of the slope difference between the global upper and lower edge baselines. The smaller the value, the more parallel the two edges are. This reflects the uniformity and accuracy of the anchor bolt width; the smaller the value, the more uniform and closer to the standard the anchor bolt width. These two parts together measure the symmetry of the anchor bolt. The first and second parts, added together, constitute a comprehensive assessment of the anchor bolt's straightness. The smaller the value, the straighter and more standard the shape of the anchor bolt.

[0091] Thus, the straightness that can quantitatively describe the degree of bending of the anchor rod under test has been obtained.

[0092] Step S600: Determine the safety status of the anchor rod to be tested based on the straightness of the global edge baseline.

[0093] It should be noted that the calculated straightness is a continuous value. To make it clearly instructive in industrial production, it needs to be mapped to discrete safety levels. This allows the system to automatically make judgments based on the quantification results and trigger corresponding subsequent operations, thereby achieving closed-loop control of the production process.

[0094] Specifically, based on production process requirements and quality control standards, the straightness index is... Set grading thresholds. For example, the safety status of anchor bolts can be divided into three levels:

[0095] Safety: When At that time, it is determined that the anchor bolt is in good condition and meets the factory standards. For example, it can be set to... Can be set to .

[0096] Warning: When At that time, it was determined that the anchor bolt had slight deformation, which was acceptable but required attention. The system can issue a prompt message to the operator, suggesting that they check the straightening machine and other related equipment. For example, it can be set... for .

[0097] Danger: When If the anchor bolt is found to be severely bent, it is deemed a defective product. The system should immediately trigger an audible and visual alarm and can also link with the control system to automatically remove the anchor bolt from the production line to prevent it from flowing into the next process.

[0098] It should be added that, and It is not fixed, but can be flexibly configured by the implementers according to different specifications of anchor bolts and specific quality requirements.

[0099] The steps above used It is a standard normalization function used to quantize calculation results to... For the interval, specific methods such as minimum-maximum normalization and Z-score standardization can be used, which are all existing technologies and will not be elaborated on here.

[0100] The second aspect of this embodiment provides a visual monitoring system for the safe production of mine anchor bolts, such as... Figure 6 As shown, the visualization monitoring system for safe production of mine anchor bolts includes a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the first aspect of the present invention, a visualization monitoring method for safe production of mine anchor bolts, is implemented.

[0101] The visualization monitoring system for safe production of mining anchor bolts also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0102] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (DRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (DRAM), high-bandwidth memory, hybrid memory cube, etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.

[0103] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A visual monitoring method for safe production of mine anchor bolts, characterized in that, Including the following steps: Acquire the image of the anchor rod to be tested, and perform edge detection on the image of the rod to obtain the edge image of the anchor rod to be tested; Multi-scale decomposition of the edge image yields scaled images at several scales; For each scale of the image, line detection is performed to extract multiple line segments, and fitting is performed based on the multiple line segments to obtain the edge baseline at the corresponding scale. We obtain the global edge baseline by weighted fusion of edge baselines at all scales. The weights of the weighted fusion are positively correlated with the product of the gradient consistency and edge sharpness of the pixels on the edge baseline at the corresponding scale, and are also positively correlated with the scale size. The acquisition of gradient consistency includes: calculating the sum of squares of the differences between the local gradients of all pixels on the edge baseline at the same scale and the average gradient of the edge baseline, and recording the result of negative correlation normalization of the sum of squares as gradient consistency. Edge sharpness satisfies the following relationship: ;in, It is the first The first layer at the layer scale Edge sharpness of the baseline; It is the first The first layer at the layer scale The number of valid straight line segments of the edge baseline; This is the preset maximum number of valid straight line segments; It is the first Layer scale The average gradient magnitude of all pixels along the edge baseline; It is the average gradient magnitude of preset pixels; calculates the straightness of the global edge baseline; the straightness is positively correlated with the average distance from the global edge baseline to the preset qualified anchor edge, the slope difference of the global edge baseline, and the average spacing; The safety status of the anchor rod to be tested is determined based on the straightness of the global edge baseline.

2. The visual monitoring method for safe production of mine anchor bolts according to claim 1, characterized in that, The fitting based on the multiple line segments to obtain the edge baseline at the corresponding scale includes: Based on the angle difference between each straight line segment and the main axis direction of the anchor rod to be tested, and the ratio of its length to the average length of all straight lines under the current scale, the probability that each straight line segment belongs to a valid straight line segment is calculated. Line segments with a probability greater than a preset probability threshold are identified as valid line segments, and the edge baseline is obtained by fitting based on the valid line segments.

3. The visual monitoring method for safe production of mine anchor bolts according to claim 2, characterized in that, The probability that the line segment is a valid line segment Satisfying the relation: ;in, It is the first at the current scale The angle between the straight line segment and the main axis direction of the anchor bolt; It is the direction of the anchor bolt's main axis; It is the first at the current scale The length of a straight line segment; It is the average length of all straight line segments at the current scale; It is a natural exponential function; It is a standard normalized function; It is the absolute value symbol.

4. The visual monitoring method for safe production of mine anchor bolts according to claim 1, characterized in that, The step of weighted fusing edge baselines at all scales to obtain a global edge baseline includes: The edge reference line at each scale includes an upper edge reference line and a lower edge reference line; For pixels on the top edge baseline at all scales, the global top edge baseline is obtained by fitting using the least squares method based on the weighted fusion weights. For pixels on the lower edge baseline at all scales, the global lower edge baseline is obtained by fitting using the least squares method based on the weighted fusion weights; The global upper edge baseline and the global lower edge baseline together constitute the global edge baseline.

5. The visual monitoring method for safe production of mine anchor bolts according to claim 4, characterized in that, Straightness of the global edge baseline Satisfying the relation: ； in, , These are the average distances from all pixels on the global top and bottom edge baselines to the edge of the preset qualified anchor bar; , These are the slopes of the global upper and lower edge baselines, respectively; It is the average distance between the global upper and lower edge baselines; It is the average distance between the upper and lower edge reference lines of the pre-set qualified anchor rod; It is a standard normalized function; It is the absolute value symbol.

6. The visual monitoring method for safe production of mine anchor bolts according to claim 1, characterized in that, The multi-scale decomposition of the edge image is achieved through the Laplacian pyramid decomposition algorithm.

7. The visual monitoring method for safe production of mine anchor bolts according to claim 1, characterized in that, The line detection is achieved using the LSD line detection algorithm.

8. A visual monitoring system for safe production of mine anchor bolts, characterized in that, The visualization monitoring system for safe production of mining anchor bolts includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a visualization monitoring method for safe production of mining anchor bolts according to any one of claims 1-7 is implemented.

Images