A quality detection system for size of sandstone crushing granularity

Abstract

The application discloses a kind of quality detection systems of sandstone crushing granularity size, it is related to the technical field of gravel detection, the system can efficiently, accurately obtain the multi-angle image data of sandstone particle by multi-angle image acquisition module and three-dimensional reconstruction module, and high-precision three-dimensional particle model is generated by image stitching and three-dimensional reconstruction technology.Through the analysis of the triangular mesh vertex vector of three-dimensional model, the equivalent particle diameter and uniformity index of particle are calculated, and the automatic evaluation of the quality of sandstone particle is realized.The quality and uniformity of sandstone particle can be accurately evaluated, accurate data support is provided for quality control in production process, and according to the size and uniformity of sandstone particle, crushing process adjustment early warning is generated in time, and accurate quality detection data is provided for production line.This technology significantly improves the detection accuracy, replaces the traditional manual screening method, reduces human error, improves production efficiency and automation degree.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sand and gravel detection, in particular to a quality detection system for the particle size of crushed sand and gravel. BACKGROUND

[0002] Crushed stone refers to small pieces of broken rock, also known as sand and gravel, which has irregular size, shape, and texture, and is usually obtained by crushing large rocks. Rock is a solid aggregate with a stable shape composed of one or more minerals and natural glass. Rock composed of one mineral is called monomineralic rock, such as marble composed of calcite and quartz rock composed of quartz. Currently, sand and gravel crushing technology is widely used in construction, mining, metallurgy, and road construction industries, and plays a crucial role in the deep processing of sand and gravel resources. With the increasing demand for sand and gravel particle quality, how to accurately and quickly detect the particle size, uniformity, and crushing effect of crushed stone has become a major challenge in the sand and gravel production process. Meanwhile, in traditional quality detection methods, particle size usually relies on manual screening or two-dimensional image analysis based on a single angle, which is not only time-consuming but also has low accuracy. Therefore, how to comprehensively and carefully evaluate the particle size of sand and gravel particles through advanced image processing technology has become a focus of current technical research.

[0003] Although existing particle size detection systems use visual sensors or simple image recognition methods to screen and analyze sand and gravel particles, these methods often exhibit low accuracy and stability in cases where particle morphology is complex and particle size varies greatly. Traditional two-dimensional images can only capture the projection information of particles from one angle, making it difficult to fully reflect the true three-dimensional morphology of particles, especially when particles overlap, are obscured, or have complex textures on the surface. In addition, these technologies often ignore the morphological factors of particles, making it impossible to accurately measure volume and reconstruct three-dimensional models, resulting in large errors and uncertainties in particle size evaluation.

[0004] Therefore, there is an urgent need for a technical solution that can accurately analyze sand and gravel particles from multiple angles and dimensions to improve detection accuracy and adapt to different production scenarios. SUMMARY

[0005] To address the problems mentioned in the background art, the present application provides a quality detection system for the particle size of crushed sand and gravel.

[0006] To achieve the above purpose, the present application provides a quality detection system for the particle size of crushed sand and gravel, comprising:

[0007] The multi-angle acquisition module, the image preprocessing module, the image splicing and three-dimensional reconstruction module, the particle quality detection module and the screen state evaluation module;

[0008] The multi-angle acquisition module is installed above the conveying belt for transporting the sandstone to be measured, and acquires multi-angle sandstone particle image data sets by continuously shooting the sandstone particles on the conveying belt according to the conveying speed of the conveying belt and the camera frame rate, and transmits the multi-angle sandstone particle image data sets to the image preprocessing module;

[0009] The image preprocessing module is used for preprocessing the multi-angle sandstone particle image data sets, which includes blur detection and elimination, depth analysis and denoising processing, and outputs a standard sandstone particle image set;

[0010] The image splicing and three-dimensional reconstruction module is used for extracting feature point descriptors of all images in the standard sandstone particle image set, matching the feature point descriptors to obtain a feature point matching value, judging the matching condition based on the output result of the feature point matching value, registering multi-angle sandstone particle images based on the feature point descriptors to generate a sparse point cloud model, and expanding the sparse point cloud model into a triangular mesh structure to restore a three-dimensional particle model of the sandstone particle;

[0011] The particle quality detection module is used for obtaining a triangular mesh vertex vector of the triangular mesh structure, calculating an equivalent particle diameter of the sandstone particle according to the triangular mesh vertex vector, calculating a uniformity index of the sandstone particle based on the equivalent particle diameter, and preliminarily comparing and evaluating the uniformity index with a preset crushing standard threshold value;

[0012] The screen state evaluation module includes an optical sensor arranged at the entrance and exit of the screen, and is used for counting screening data of the sandstone particles, calculating a passing rate of the screen based on the screening data, and secondarily comparing and evaluating the passing rate with a set passing threshold value to obtain a damage condition of the screen.

[0013] Preferably, the multi-angle acquisition module includes an acquisition setting unit, a data processing unit and an image data transmission unit;

[0014] The acquisition setting unit includes a conveying belt speed sensor and a plurality of camera groups, which are respectively installed directly above the conveying belt and on both sides of the conveying belt, and the conveying belt speed sensor is used for acquiring the conveying speed of the conveying belt;

[0015] The data processing unit is used to calculate the camera frame rate of the camera group, and emit auxiliary line frames through a laser marker to lock the sand and gravel particle area. The camera group continuously captures multi-angle image data of the same sand and gravel particle on the conveying belt through frame synchronization technology at the camera frame rate F, and records the number of the sand and gravel particle and the time stamp of the shooting, to obtain a multi-angle sand and gravel particle image data set.

[0016] The image data transmission unit is used to transmit the multi-angle sand and gravel particle image data set collected by the camera group to the image preprocessing module.

[0017] Preferably, the image preprocessing module comprises a blur detection and elimination unit, a depth analysis unit and a denoising processing unit.

[0018] The blur detection and elimination unit is used to receive the collected multi-angle sand and gravel image data, calculate the sharpness index corresponding to each sand and gravel image in the multi-angle sand and gravel image data respectively by using the Laplacian gradient method, and compare the sharpness index with a set blur threshold value:

[0019] When the sharpness index > the blur threshold value, it is determined as a first sand and gravel image, which is a clear sand and gravel image, and is retained. When the sharpness index ≤ the blur threshold value, it is determined as a second sand and gravel image, which is a blurred sand and gravel image, and is eliminated.

[0020] The depth analysis unit is used to analyze the first sand and gravel image through a barrage deep learning algorithm, determine the distance from the pixels of the first sand and gravel image to the camera, and generate a depth map and a depth value corresponding to each pixel point coordinate.

[0021] The denoising processing unit is used to perform refined denoising processing on the boundary of the first sand and gravel image through a convolutional neural network (CNN), optimize the image sharpness, and after the denoising processing, the remaining sand and gravel images are summarized to output a standard sand and gravel particle image set.

[0022] Preferably, the image stitching and three-dimensional reconstruction module comprises a feature point extraction unit and a feature point matching unit.

[0023] The feature point extraction unit is used to perform Gaussian blur of different scales on all images in the standard sand and gravel particle image set, generate a multi-layer pyramid to detect features of different sizes, perform difference calculation on the pyramid to obtain image differences between adjacent scales, extract extreme points as preliminary feature points, determine the accurate position and main direction of the feature points, give rotation invariance, calculate the gradient direction histogram in the neighborhood of the feature points, and form a feature point descriptor.

[0024] The feature point matching unit is configured to perform feature point matching on the feature point descriptors in combination with the depth values to obtain feature point matching values, and to determine the overlapping area of the multi-angle sandstone images in the standard sandstone particle image set based on the output results of the feature point matching values, align the two images by matching the feature points, and generate a two-dimensional sandstone image.

[0025] The determination based on the output results of the feature point matching values is as follows.

[0026] When the feature point matching value = 1, the two feature points are matching points.

[0027] When the feature point matching value = 0, the two feature points are not matching points.

[0028] Preferably, the image stitching and three-dimensional reconstruction module further comprises a point cloud registration and stitching unit.

[0029] The point cloud registration and stitching unit is configured to randomly select a subset from all matching points by using the RANSAC algorithm, fit a transformation matrix model, calculate the matching point error e of all matching points to the transformation matrix model, compare the matching point error e with an error threshold eth to determine the accuracy of the matching points, and repeatedly iterate multiple times to select the transformation matrix model with an upper limit of inliers as the matching points and eliminate the error matching points.

[0030] When the matching point error < the error threshold, the point pair is considered to be successfully matched, and is marked as an inlier.

[0031] When the matching point error ≥ the error threshold, the point pair is considered to be unsuccessfully matched, and the matching points are eliminated.

[0032] Preferably, the image stitching and three-dimensional reconstruction module further comprises a matching point elimination unit. The point cloud registration and stitching unit performs point cloud registration and stitching by applying the ICP algorithm, forms a sparse point cloud model of the complete sandstone particles, performs a near-neighbor search on the sparse point cloud model, constructs a triangular mesh connection of the points, optimizes the triangular mesh to complete the boundary, generates a closed surface, and restores a three-dimensional particle model of the sandstone particles.

[0033] The triangular mesh determines the position coordinates of each vertex in the three-dimensional space through three vertex vectors.

[0034] Preferably, the particle quality detection module comprises a particle size analysis unit, a uniformity analysis unit, and a crushing effect evaluation unit. The particle size analysis unit extracts the vertex coordinates of each triangular mesh from the three-dimensional particle model and generates three vertex vectors of each triangular mesh.

[0035] Based on summation of all vertex vectors of the triangular mesh, a closed three-dimensional volume defined by the triangular mesh is calculated, the closed three-dimensional volume is the volume of the sand and stone particles, and an equivalent particle diameter of the sand and stone particles is calculated as an output.

[0036] Preferably, the uniformity analysis unit determines the uniformity index of the sand and stone particles by calculating the equivalent particle diameters of all three-dimensional particle models and performing mean value and variance calculation on all the equivalent particle diameters, and analyzes the sand and stone particle crushing process.

[0037] Preferably, the crushing effect evaluation unit sets a crushing standard threshold based on a sand and stone crushing standard, preliminarily compares and evaluates the uniformity index, analyzes the uniformity of the sand and stone particles, and determines the crushing process adjustment, and the specific evaluation content is as follows.

[0038] When the uniformity index is greater than the crushing standard threshold, it indicates that the size distribution of the sand and stone particles is uneven, and a crushing process adjustment warning is generated at this time.

[0039] When the uniformity index is less than or equal to the crushing standard threshold, it indicates that the size distribution of the sand and stone particles is uniform, and no adjustment is required at this time.

[0040] Preferably, the screen state evaluation module comprises a screening rate analysis unit and a screen state evaluation unit.

[0041] The screening rate analysis unit installs optical sensors at the outlet and inlet of the screen, and when it is preliminarily evaluated that the size distribution of the sand and stone particles is uniform, the optical sensors are started to statistically count the number of screened particles and the number of entering particles in real time, and the number of screened particles and the number of entering particles are calculated to obtain the passing rate of the screen. The screen state evaluation unit sets a passing threshold according to the production process requirements by the user, and secondarily compares and evaluates the passing rate to analyze the damage of the screen, and the specific evaluation content is as follows.

[0042] When the passing rate is greater than the passing threshold, it indicates that the screening effect of the screen is normal, and no operation is required.

[0043] When the passing rate is greater than twice the passing threshold, it indicates that the screening effect of the screen is abnormal, and it is judged that the screen is damaged, and a first screen warning is generated to prompt the replacement of the screen.

[0044] When the passing rate is less than or equal to the passing threshold, and the time is greater than 5 seconds, it indicates that the screening effect of the screen is abnormal, and it is judged that the surface of the screen is blocked, and a second screen warning is generated to prompt the cleaning of the surface of the screen.

[0045] The present application has the following beneficial effects:

[0046] (1) The system can efficiently and accurately obtain multi-angle image data of sandstone particles through the multi-angle image acquisition module and the three-dimensional reconstruction module, and generate a high-precision three-dimensional particle model through image stitching and three-dimensional reconstruction technology. By analyzing the triangular mesh vertex vectors of the three-dimensional model, the equivalent particle size and uniformity index of the particle are calculated, and the automatic evaluation of the quality of the sandstone particle is realized. The system can generate a crushing process adjustment warning in a timely manner according to the size and uniformity of the sandstone particle, and provide accurate quality detection data for the production line. This technology significantly improves the detection accuracy, replaces the traditional manual screening method, reduces human error, and improves production efficiency and automation.

[0047] (2) The system can automatically calculate the passing rate of the screen through real-time monitoring of the screening process of the sandstone particle by the screen state evaluation module combined with the optical sensor. By setting a reasonable screen passing rate threshold F2, the system can automatically perform secondary evaluation when the screen is damaged or clogged, and generate a related warning. Specifically, when the screen passing rate exceeds twice the threshold, the system will trigger the first screen warning, prompting that the screen may have been damaged and needs to be replaced; when the passing rate is lower than the threshold and the time exceeds 5 seconds, the system will generate the second screen warning, prompting that the screen surface is clogged and needs to be cleaned. This warning mechanism can effectively reduce the production stagnation caused by screen damage or clogging, ensure the continuity and efficiency of the screening process, avoid material waste, and improve the stability of the production line.

[0048] (3) The particle size analysis unit and the uniformity analysis unit in the particle quality detection module can calculate the equivalent particle size and evaluate the uniformity of the particle based on real-time acquisition of the size data of the sandstone particle, and obtain the uniformity index. If the uniformity index exceeds the preset crushing standard threshold, the system will automatically trigger a crushing process adjustment warning to remind the operator to adjust the crushing process parameters, thereby improving the uniformity of the particle and avoiding low production efficiency caused by uneven particle size distribution. Through this intelligent evaluation and adjustment mechanism, the system can effectively optimize the sandstone crushing process, reduce particles that do not meet the specifications, improve the crushing efficiency, save energy, reduce waste, and ultimately improve the economic benefits of production. BRIEF DESCRIPTION OF DRAWINGS

[0049] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which similar reference characters refer to similar elements throughout the several views.

[0050] Figure 1 is a structural schematic diagram of the quality detection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0051] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0052] Embodiment 1

[0053] Please refer to Figure 1 The present application provides a quality detection system for particle size of sandstone crushing, in order to achieve the above purpose, the present application is realized by the following technical solutions: including multi-angle acquisition module, image preprocessing module, image stitching and three-dimensional reconstruction module, particle quality detection module and screen state evaluation module;

[0054] The multi-angle acquisition module calculates the camera frame rate F by installing a camera group on the conveyor belt in combination with the conveyor belt speed Vb, and continuously photographs the sandstone particles on the conveying belt according to the camera frame rate F through a flip-flop to obtain multi-angle sandstone particle image data sets, and constructs a quality detection system, and transmits the multi-angle sandstone particle image data sets to the quality detection system;

[0055] The image preprocessing module preprocesses the multi-angle sandstone particle image data sets in the quality detection system, which includes blur detection and rejection, depth analysis and denoising, and outputs a standard sandstone particle image set;

[0056] The image stitching and three-dimensional reconstruction module extracts the feature point descriptor f(pi) of all images in the standard sandstone particle image set, matches the feature points pi, obtains the feature point matching value Match(pi, pj), and judges the matching condition based on the output result of the feature point matching value Match(pi, pj);

[0057] At the same time, the feature point descriptor f(pi) is used to register multi-angle sandstone particle images to generate a sparse point cloud model, and the sparse point cloud model is expanded to a triangular mesh structure to restore a three-dimensional particle model of the sandstone particles;

[0058] The particle quality detection module calculates the equivalent particle diameter Deq of the sandstone particles based on the triangular mesh vertex vector of the three-dimensional particle model, and calculates the uniformity index CV of the sandstone particles based on the equivalent particle diameter Deq, and then preliminarily compares and evaluates the crushing standard threshold F1 and the uniformity index CV;

[0059] The screen state evaluation module installs optical sensors at the entrances and exits of the screen, counts the screening data of the sand and stone particles, calculates the passing rate R of the screen based on the screening data, and sets a passing threshold F2 for secondary comparison and evaluation with the passing rate R to analyze the screen damage.

[0060] In this embodiment, the multi-angle acquisition module of the system installs a camera group on the conveyor belt and calculates the camera frame rate F in combination with the speed Vb of the conveyor belt. Under the control of the trigger, the sand and stone particles are continuously photographed at multiple angles to obtain detailed particle image data sets. Then, the image is processed by the image preprocessing module, including blur detection, depth analysis and denoising, to output standardized particle image sets, ensuring the quality and accuracy of the image data. Next, the image stitching and three-dimensional reconstruction module extracts feature points from the image, matches and registers them, generates a sparse point cloud model, and further expands it into a triangular mesh structure to restore the three-dimensional morphology of the sand and stone particles. On this basis, the particle quality detection module calculates the equivalent particle size Deq using the triangular mesh vertex vector of the three-dimensional particle model, and evaluates the crushing effect according to the uniformity index CV of the particles. When the uniformity index CV of the particles is higher than the set threshold, the system will generate a crushing process adjustment warning to prompt the operator to optimize and adjust the crushing process, ensuring the uniformity of the particle size and the crushing efficiency. At the same time, the screen state evaluation module monitors the screening situation in real time in combination with the optical sensor, calculates the passing rate R of the screen, and further evaluates the working state of the screen. When the passing rate is lower than the set threshold, the system will automatically trigger a screen damage or clogging warning to prompt the operator to maintain or replace the screen. Through the cooperative work of these modules, the quality detection system of the present application not only greatly improves the precision and automation of sand and stone particle detection, but also optimizes the crushing process in real time and improves the overall efficiency of the production line. Specifically, it can accurately evaluate the quality and screening effect of the particles, reduce production downtime caused by uneven particle size or screen problems, optimize the production process, reduce energy consumption and material waste, and ultimately improve the stability and production efficiency of the sand and stone crushing process.

[0061] Embodiment 2

[0062] This embodiment is an explanation and description in Embodiment 1, please refer to Figure 1 , specifically: the multi-angle acquisition module includes an acquisition setting unit and an image data transmission unit;

[0063] The acquisition setting unit installs a camera group above and on both sides of the sand and stone conveying belt, and uses a conveying belt speed sensor to obtain the conveying belt speed Vb to calculate the camera frame rate F. The algorithm formula of the camera frame rate F is: , wherein Dmin represents the minimum diameter of the sand and stone particles, and S represents the safety factor; the laser marker is used to emit an auxiliary frame line to lock the sand and stone particle area; the camera continuously captures the multi-angle image data of the same sand and stone particle on the conveying belt at the camera frame rate F through frame synchronization technology; the camera automatically records the number of the sand and stone particle and the time stamp of the captured image, and the multi-angle sand and stone particle image data set is obtained;

[0064] The image data transmission unit directly transmits the multi-angle sand and stone particle image data set collected by the camera group to the quality detection system through the line connection between the camera group and the quality detection system.

[0065] In this embodiment, the multi-angle acquisition module of the system realizes efficient and multi-angle real-time image acquisition and transmission of the sand and stone particles through the close cooperation of the acquisition setting unit and the image data transmission unit. Specifically, the acquisition setting unit installs a plurality of camera groups above and on both sides of the sand and stone conveying belt, combines the conveying belt speed sensor to obtain the conveying belt speed Vb in real time, and calculates the frame rate F of the camera according to the speed. The camera frame synchronization technology and the laser marker auxiliary frame line are used to lock the particle area to ensure that the camera can continuously capture the sand and stone particle image on the conveying belt at the correct frame rate F, and automatically record the number and time stamp of each particle to capture the multi-angle sand and stone particle image data set. These image data are directly transmitted to the quality detection system through the image data transmission unit, realizing fast and stable data transmission. Through this design, the multi-angle image data of the sand and stone particles can be efficiently and accurately acquired, providing a high-quality image data basis for subsequent particle quality analysis, three-dimensional reconstruction, and screen state evaluation.

[0066] Embodiment 3

[0067] This embodiment is an explanation and description in embodiment 2. Please refer to Figure 1 Specifically, the image preprocessing module includes a blur detection and elimination unit, a depth analysis unit, and a denoising processing unit.

[0068] The blur detection and elimination unit receives the collected multi-angle sand and stone image data in real time in the quality detection system, calculates the sharpness index L of the sand and stone image in the multi-angle sand and stone image data by using the Laplacian gradient method, sets a blur threshold Lmin, and performs blur detection and elimination evaluation. The specific evaluation content is as follows:

[0069] When the sharpness index is greater than the blur threshold, it is determined as a first sand and stone image, which is a clear sand and stone image, and is retained. When the sharpness index is less than or equal to the blur threshold, it is determined as a second sand and stone image, which is a blurred sand and stone image, and is eliminated.

[0070] The depth analysis unit is configured to analyze the first sandstone image by a barrage depth learning algorithm, determine the distance from the pixels of the first sandstone image to the camera, and generate a depth map and a depth value corresponding to each pixel point coordinate;

[0071] The specific algorithm formula of the clarity index L is: ; wherein, represents a partial derivative, I(x, y) represents the pixel value of the image at the pixel point coordinate (x, y), x represents the horizontal axis direction, and y represents the vertical axis direction;

[0072] The depth analysis unit analyzes the distance from the pixels of each sandstone image in the multi-angle sandstone image data to the camera by the barrage depth learning algorithm after removing the blurred images, and generates a depth map D(x, y) and a depth value Z(x, y) corresponding to each pixel point coordinate (x, y).

[0073] The denoising processing unit refines and denoises the boundaries of each sandstone image in the multi-angle sandstone image data by introducing a convolutional neural network CNN, optimizes the image clarity, and after denoising, the remaining sandstone images are summarized and a standard sandstone particle image set is output.

[0074] In this embodiment, the image preprocessing module of the system cooperates with sub-modules such as the blur detection and removal unit, the depth analysis unit, and the denoising processing unit, significantly improving the quality and reliability of the sandstone particle image data and ensuring the accuracy of the subsequent image stitching and three-dimensional reconstruction process. The specific implementation is that the blur detection and removal unit first calculates the clarity index L of the sandstone image using the Laplacian gradient method, sets a blur threshold Lmin, and evaluates the clarity of each image in real time. When the clarity index L of the image is greater than the threshold, the image is retained; if L is less than or equal to the threshold, it is determined as a blurred image and is removed, reducing the interference of blurred images on subsequent analysis results. Through this preprocessing step, low-quality image data can be effectively removed, thereby improving the overall quality of the image data set. Next, the depth analysis unit analyzes the distance between the sandstone particles and the camera in each image by a depth learning algorithm, generates an accurate depth map D(x, y) and a depth value Z(x, y) for each pixel point, and provides spatial position information for subsequent three-dimensional reconstruction and particle size analysis. This depth analysis process ensures the stereoscopic and realistic feeling of the image data, making the spatial distribution and morphological characteristics of the particles more accurate. Finally, the denoising processing unit uses a convolutional neural network CNN to refine and denoise the boundaries of the image, further optimizing the clarity of the image and removing background noise and irrelevant information. After denoising, the system will obtain a set of standardized, clear and accurate sandstone particle images, providing high-quality input data for subsequent image stitching, three-dimensional reconstruction and particle quality detection.

[0075] Embodiment 4

[0076] This embodiment is an explanation in embodiment 3, please refer to Figure 1 , specifically: image stitching and three-dimensional reconstruction module includes feature point extraction and matching unit and registration stitching unit;

[0077] Feature point extraction and matching unit includes feature point extraction unit and feature point matching unit;

[0078] Feature point extraction unit generates multi-layer pyramid by Gaussian blur of different scales on all images in standard sandstone particle image set, detects different size features, and extracts extreme points as preliminary feature points by difference calculation of adjacent scales of the pyramid, determines the accurate position and main direction of the feature points, gives rotation invariance, calculates the gradient direction histogram in the neighborhood of the feature points, forms feature point descriptor f(pi), feature point descriptor f(pi)={position, scale, direction, descriptor};

[0079] Feature point matching unit matches feature points by feature point descriptor f(pi) of all sandstone images in standard sandstone particle image set, combines depth value Z(x,y), obtains feature point matching value Match(pi,pj), and judges matching condition based on output result of feature point matching value Match(pi,pj), finds out overlapping area of multi-angle sandstone images in standard sandstone particle image set, aligns two images by matching feature points, and generates two-dimensional sandstone image;

[0080] The judgment formula of feature point matching value Match(pi,pj) is:

[0081] ;

[0082] In the formula, otherwise means otherwise, if means condition, f(pj) means descriptor of feature point pj, z(x,y) pi means depth value of feature point pi, z(x,y) pj means depth value of feature point pj, means depth value weight factor, used to balance the influence of feature descriptor similarity and depth difference, When the depth difference is larger, the importance of matching is higher; When the depth difference is smaller, it mainly depends on the similarity of the descriptor, Means matching threshold;

[0083] The specific judgment is as follows:

[0084] The feature point matching value Match (pi, pj) = 1 indicates that the i-th feature point pi and the j-th feature point pj are matching points.

[0085] The feature point matching value Match (pi, pj) = 0 indicates that the i-th feature point pi and the j-th feature point pj are not matching points.

[0086] The registration splicing unit comprises a matching point elimination unit and a point cloud registration splicing unit.

[0087] The point cloud registration splicing unit selects a subset from the matching points at random by using the RANSAC algorithm, fits a transformation matrix model, calculates the matching point error e of all matching points to the transformation matrix model, compares the matching point error e with an error threshold eth to judge the accuracy of the matching points, and repeatedly iterates several times to select the transformation matrix model with an upper limit of inliers as the matching points and eliminate the error matching points.

[0088] The specific algorithm formula of the matching point error e is: wherein H represents the transformation matrix.

[0089] The specific judgment is as follows.

[0090] When the matching point error e is less than the error threshold eth, the point pair is considered to be matched successfully, and (pi, pj) is an inlier.

[0091] When the matching point error e is greater than or equal to the error threshold eth, the point pair is considered to be matched unsuccessfully, and (pi, pj) is eliminated.

[0092] The error threshold eth is obtained by experiments to obtain a typical matching error range, then a threshold value capable of distinguishing correct matching and error matching is selected, and then an initial setting is made by relevant staff.

[0093] The point cloud registration splicing unit performs point cloud registration splicing by applying the ICP algorithm, forms a sparse point cloud model of complete sand particles, performs a near neighbor search on the sparse point cloud model, constructs a triangular mesh connection of points, optimizes the triangular mesh to complete the boundary, generates a closed surface, and restores a three-dimensional particle model of the sand particles.

[0094] The triangular mesh is determined by three vertex vectors The position coordinates (x v , y v , z v ) of each vertex contained in the three-dimensional space are determined, which represent three vertices of a triangle and define the position of the triangle in the three-dimensional space.

[0095] In this embodiment, the feature point extraction and matching unit of the system extracts feature points in the sand and gravel particle image through Gaussian blur and difference pyramid algorithm, and generates a rotation-invariant descriptor f(pi) for each feature point. This process ensures that the feature points have strong robustness and can effectively identify the same particles under different angles. Then, the feature point matching unit combines the depth information of the image, compares the depth value and the feature point descriptor, accurately matches the feature points of the same particle in different images, and finds the overlapping area between the images. This matching process judges the correctness of the matching through the feature point matching value Match(pi,pj), ensuring high quality and low error of the matching points.

[0096] On the basis of feature point matching, the registration and splicing unit selects inliers from the matching points through the RANSAC algorithm and fits a transformation matrix model to further eliminate false matching points. Through multiple iterations, the system can accurately calculate the transformation relationship of the matching points and combine the sand and gravel particle information in multiple images into a complete sparse point cloud model through point cloud registration and splicing technology. Then, based on the ICP algorithm, the registration effect of the point cloud is further optimized, the points are connected through triangular mesh, the boundary is completed, and a closed surface is generated, successfully restoring the three-dimensional morphology of the particles. This process generates a high-precision three-dimensional particle model for each sand and gravel particle through accurate image splicing and three-dimensional reconstruction, significantly improving the accuracy of particle size measurement and quality detection. At the same time, the three-dimensional reconstruction result provides more accurate particle distribution information for screen state evaluation, thereby improving the evaluation accuracy of the screening effect.

[0097] Embodiment 5

[0098] This embodiment is an explanation and description in Embodiment 4, please refer to Figure 1 , in particular: the particle quality detection module includes a particle size analysis unit, a uniformity analysis unit and a crushing effect evaluation unit;

[0099] The particle size analysis unit extracts the vertex coordinates (x, y, z) of each triangular mesh from the three-dimensional particle model, calculates three vertex vectors for each triangular mesh, and the specific algorithm is as follows:

[0100] ;

[0101] The closed three-dimensional volume defined by the triangular mesh is calculated by summing all the vertex vectors of the triangular mesh, the volume V of the sand and gravel particle is obtained, and the volume V of the sand and gravel particle is calculated to output the equivalent particle size Deq of the sand and gravel particle;

[0102] The equivalent particle size Deq is calculated by the following algorithm formula;

[0103] ; wherein ;

[0104] where N represents the total number of triangular meshes, π represents the ratio of a circle's circumference to its diameter, and takes the value 3.14, represents the dot product of the vertex vector and the normal vector, and represents the projection of the triangular volume, represents the cross product of the two vertex vectors, and represents the normal vector of the triangle;

[0105] The equivalent diameter can be derived from the volume, and a standardized size indicator can be used to represent the particle size, even if the particle shape is irregular.

[0106] The uniformity analysis unit calculates the equivalent particle diameter Deq for all three-dimensional particle models, and calculates the average value and variance of all equivalent particle diameters Deq to obtain the uniformity index CV of the sand and gravel particles, and analyzes the sand and gravel particle crushing process;

[0107] The uniformity index CV is calculated and output by the following algorithm formula;

[0108] ;

[0109] where Deq n represents the equivalent diameter of the nth particle, represents the average value of the equivalent diameter.

[0110] The crushing effect evaluation unit sets a crushing standard threshold F1 based on the sand crushing standard, and preliminarily compares and evaluates the uniformity index CV to analyze the uniformity of the sand and gravel particles, and determines the crushing process adjustment, and the specific evaluation content is as follows;

[0111] When the uniformity index CV is greater than the crushing standard threshold F1, it indicates that the size distribution of the sand and gravel particles is uneven, and at this time, the crushing process adjustment warning is generated;

[0112] When the uniformity index CV is less than or equal to the crushing standard threshold F1, it indicates that the size distribution of the sand and gravel particles is uniform, and at this time, no adjustment is needed.

[0113] In this embodiment, the particle size analysis unit of the system extracts the vertex coordinates (x, y, z) of the triangular meshes from the three-dimensional particle model, and calculates the vertex vectors of each triangular mesh, and then uses the volume derivation formula of the triangular mesh to calculate the volume V of the sand and gravel particles. Through these volume data, the equivalent particle diameter Deq of the sand and gravel particles is further derived. The equivalent particle diameter provides a standardized size indicator, even if the sand and gravel particle shape is irregular, it can accurately describe its size. This analysis provides basic data for further quality detection of particles, ensuring high-precision measurement of particle size.

[0114] Secondly, the uniformity analysis unit calculates the average and variance of the equivalent particle diameter Deq of all sand particles to obtain the uniformity index CV. The uniformity index CV reflects the distribution of particle size, helping to analyze whether the sand particles are uniform in the crushing process. A lower CV value indicates that the particle size distribution is more uniform, while a higher CV value means that the particle distribution is uneven, which may indicate problems in the crushing process. Finally, the crushing effect evaluation unit sets a crushing standard threshold F1 and evaluates it in combination with the uniformity index CV. When the CV value exceeds the threshold F1, the system generates a crushing process adjustment warning, indicating that the particle size distribution is uneven and the crushing process may need to be adjusted. When the CV value is less than or equal to F1, it indicates that the particle size distribution is uniform, and the system does not need to be adjusted. This module not only helps monitor the uniformity of sand particles, but also provides real-time feedback on the effectiveness of the crushing process, allowing timely adjustments to the production process to ensure the stability of sand product quality.

[0115] Example 6

[0116] This example is an explanation and illustration in Example 5, please refer to Figure 1 , specifically: the screen state evaluation module includes a screening rate analysis unit and a screen state evaluation unit;

[0117] The screening rate analysis unit installs optical sensors at the outlet and inlet of the screen. When the initial evaluation of the sand particle size distribution is uniform, the optical sensors are activated to real-time count the number of screened particles Npass and the number of entering particles Ntotal, and the number of screened particles Npass and the number of entering particles Ntotal are given to calculate the passing rate R of the screen. The passing rate R is calculated as follows: ;

[0118] The screen state evaluation unit sets a passing threshold F2 according to the production process requirements by the user, and then compares and evaluates the passing rate R again to analyze the damage of the screen. The specific evaluation content is as follows:

[0119] When the passing rate R is greater than the passing threshold F2, it indicates that the screen separation effect is normal and no operation is needed.

[0120] When the passing rate R is greater than twice the passing threshold F2, it indicates that the screen separation effect passing rate R is abnormal, indicating that the screen is damaged. At this time, the first screen warning is generated to prompt the replacement of the screen.

[0121] When the passing rate R is less than or equal to the passing threshold F2 and the time is greater than 5 seconds, it indicates that the screen separation effect is abnormal, indicating that the screen surface is blocked. At this time, the second screen warning is generated to prompt the cleaning of the screen surface.

[0122] In this embodiment, the screening rate analysis unit of the system calculates the passing rate R of the screen by installing optical sensors at the inlet and outlet of the screen, and counting the number of particles passing through the screen Npass and the number of particles entering the screen Ntotal in real time. Specifically, when the system preliminarily evaluates that the size distribution of the sand and stone particles is uniform, the optical sensors are started to collect real-time data of the particles passing through the screen. The calculation formula of the passing rate R is, which can directly reflect the screening efficiency of the screen under the current operating conditions. This analysis provides timely and accurate feedback on the state of the screen. Secondly, the screen state evaluation unit sets a passing rate threshold F2 according to the user's set production process requirements, and performs secondary comparative evaluation on the actual calculated passing rate R. According to the relationship between the passing rate and the threshold, the system can accurately judge the working state of the screen: when the passing rate R is greater than the threshold F2, it indicates that the screening effect of the screen is normal, and no operation is needed. When the passing rate R is more than twice the threshold F2, it indicates that the screening effect of the screen is abnormal, and there may be a screen damage situation, at which time the system will generate a first screen warning to remind the operator to replace the screen. When the passing rate R is less than or equal to the threshold F2, and the time exceeds 5 seconds, it indicates that the surface of the screen may be blocked, and the screening effect is seriously decreased, at which time the system will generate a second screen warning to remind the operator to clean the screen. Through these fine evaluation and warning mechanisms, this module not only can monitor the screening effect of the screen in real time, but also can effectively predict and discover potential problems of screen damage or blockage in advance, so as to provide accurate guidance for screen maintenance and replacement in the production process.

[0123] While the embodiments of the present application have been shown and described, it is to be understood that the embodiments can be varied, modified, replaced and changed in many ways by those skilled in the art without departing from the principles and spirit of the present application.

Claims

1. A quality detection system for sand and gravel crushing particle size, characterized in that: include: Multi-angle acquisition module, image preprocessing module, image stitching and 3D reconstruction module, particle quality detection module and screen status assessment module; The multi-angle acquisition module is installed above a conveyor belt, and the conveyor belt is used to transport sand and gravel to be measured. The multi-angle acquisition module continuously shoots the sand and gravel particles on the conveyor belt according to the conveying speed of the conveyor belt and the camera frame rate to obtain a multi-angle sand and gravel particle image dataset, and transmits the multi-angle sand and gravel particle image dataset to the image preprocessing module; The image preprocessing module is used to preprocess the multi-angle sand and gravel particle image data set, and the preprocessing includes blur detection and elimination, depth analysis and denoising, and outputs a standard sand and gravel particle image set; The image stitching and 3D reconstruction module is used to: (a) performing multi-scale pyramidal processing on the standard sand and gravel particle image set, extracting extreme points in the image as feature points through differential calculation, and generating feature point descriptors with rotation invariance; (b) matching the feature point descriptors, wherein the feature point matching process combines the depth value, calculates the weighted sum of the feature descriptor difference and the depth difference, and compares it with the matching threshold to obtain the feature point matching value; when the weighted sum is less than the matching threshold, the matching value is 1, indicating that the feature point matching is successful; otherwise, the matching value is 0, indicating that the matching fails; (c) Using the RANSAC algorithm, randomly select a subset of matching points to fit the transformation matrix model, calculate the matching point error e from all matching points to the transformation matrix model, and compare the matching point error e with the error threshold eth: when e < eth, mark it as an inlier point, and when e ≥ eth, remove the matching point; repeat the iteration until the transformation matrix with the largest number of inliers is obtained; (d) Apply the ICP algorithm to perform point cloud registration and splicing of the internal points to generate a sparse point cloud model of sand and gravel particles; (e) performing a nearest neighbor search on the sparse point cloud model, constructing a triangular mesh to connect and complete the boundaries, and generating a three-dimensional particle model of sand and gravel particles; The particle quality detection module is used to obtain the triangular mesh vertex vectors of the triangular mesh structure, calculate the equivalent particle size of the sand and gravel particles according to the triangular mesh vertex vectors, calculate the uniformity index of the sand and gravel particles based on the equivalent particle size, and perform a preliminary comparative evaluation between a preset crushing standard threshold and the uniformity index; The screen status assessment module includes an optical sensor, which is arranged at the entrance and exit of the screen. The screen status assessment module is used to count the screening data of sand and gravel particles, and calculate the pass rate of the screen based on the screening data, and perform a secondary comparative assessment of the set pass threshold and the pass rate to obtain the damage condition of the screen.

2. A quality detection system for crushed sand and gravel particles according to claim 1, characterized in that: The multi-angle acquisition module includes an acquisition setting unit, a data processing unit and an image data transmission unit; The acquisition setting unit includes a conveyor belt speed sensor and several camera groups, which are respectively installed directly above the conveyor belt and on both sides of the conveyor belt. The conveyor belt speed sensor is used to obtain the transmission speed of the conveyor belt; The data processing unit is used to calculate the camera frame rate of the camera group and emit an auxiliary wireframe through a laser calibrator to lock the sand and gravel particle area. The camera group continuously shoots at the camera frame rate F through frame synchronization technology to capture multi-angle image data of the same sand and gravel particle on the conveyor belt, and records the sand and gravel particle number and the shooting time stamp to obtain a multi-angle sand and gravel particle image data set; The image data transmission unit is used to transmit the multi-angle sand and gravel particle image data set collected by the camera group to the image preprocessing module.

3. The quality detection system for crushed sand and gravel particles according to claim 1 is characterized in that: The image preprocessing module includes a blur detection and removal unit, a depth analysis unit and a denoising processing unit; The blur detection and elimination unit is used to receive the collected multi-angle sandstone image data, calculate the clarity index corresponding to each sandstone image in the multi-angle sandstone image data by using the Laplace gradient method, and compare the clarity index with the set blur threshold: When the clarity index is greater than the blur threshold, it is determined to be the first sandstone image and retained, and the first sandstone image is a clear sandstone image; when the clarity index is less than or equal to the blur threshold, it is determined to be the second sandstone image and removed, and the second sandstone image is a blurred sandstone image; The depth analysis unit is used to analyze the first sand and gravel image using a bullet screen deep learning algorithm, determine the distance between the pixels of the first sand and gravel image and the camera, and generate a depth map and a depth value corresponding to the coordinates of each pixel point; The denoising processing unit is used to refine and denoise the boundaries of the first sand and gravel image through a convolutional neural network (CNN) to optimize image clarity. After denoising, the remaining sand and gravel images are aggregated to output a standard sand and gravel particle image set.

4. The quality detection system for crushed sand and gravel particles according to claim 3 is characterized by: The image stitching and 3D reconstruction module includes a feature point extraction unit and a feature point matching unit; The feature point extraction unit is used to perform Gaussian blurring of different scales on all images in the standard sand and gravel particle image set, generate a multi-layer pyramid to detect features of different sizes, then perform difference calculation on the pyramid to calculate image differences between adjacent scales, extract extreme points as preliminary feature points, determine the precise position and main direction of the feature points, assign rotation invariance, calculate the gradient direction histogram within the neighborhood of the feature points, and form a feature point descriptor; The feature point matching unit is used to perform feature point matching on the feature point descriptor in combination with the depth value to obtain a feature point matching value, and to make a judgment based on an output result of the feature point matching value to determine an overlapping area of ​​the multi-angle sand and gravel images in the standard sand and gravel particle image set, and to align the two images by matching feature points to generate a two-dimensional sand and gravel image; Wherein, the judgment is performed based on the output result of the feature point matching value as follows; When the feature point matching value = 1, the two feature points are matching points; When the feature point matching value = 0, the two feature points are not matching points.

5. The quality detection system for crushed sand and gravel particles according to claim 3 is characterized in that: The image stitching and 3D reconstruction module also includes a point cloud registration and stitching unit; The point cloud registration and stitching unit is used to randomly select a subset from all matching points using the RANSAC algorithm, fit the transformation matrix model, and then calculate the matching point error e from all matching points to the transformation matrix model. The matching point accuracy is determined based on the comparison of the matching point error e with the error threshold eth. The process is repeated multiple times to select the transformation matrix model with the upper limit of the inliers as the matching point and eliminate the incorrect matching points. The specific judgment is as follows; When the matching point error is less than the error threshold, the point pair is considered to be matched successfully and marked as an inlier; When the matching point error is greater than or equal to the error threshold, the point pair is considered to have failed to match and the matching point is removed.

6. A quality detection system for crushed sand and gravel particles according to claim 5, characterized in that: The image stitching and 3D reconstruction module also includes a matching point elimination unit. The point cloud registration and stitching unit performs point cloud registration and stitching by applying an ICP algorithm to form a sparse point cloud model of complete sand and gravel particles. Then, a nearest neighbor search is performed on the sparse point cloud model to construct a triangular mesh connection of the points, optimize the triangular mesh to complete the boundary, generate a closed surface, and restore the 3D particle model of the sand and gravel particles. The triangle mesh determines the position coordinates of each vertex in the three-dimensional space through three vertex vectors.

7. The quality detection system for crushed sand and gravel particles according to claim 1 is characterized in that: The particle quality detection module includes a particle size analysis unit, a uniformity analysis unit, and a crushing effect evaluation unit; the particle size analysis unit extracts the vertex coordinates of each triangular mesh from the three-dimensional particle model and generates three vertex vectors of each triangular mesh; Based on the sum of the vertex vectors of all triangular meshes, the closed three-dimensional volume defined by the triangular meshes is calculated. The closed three-dimensional volume is the volume of the sand and gravel particles, and the equivalent particle size of the sand and gravel particles is calculated and output.

8. The quality detection system for crushed sand and gravel particles according to claim 7, characterized in that: The uniformity analysis unit calculates the equivalent particle size for all three-dimensional particle models, and calculates the average value and variance of all the equivalent particle sizes to determine the uniformity index of the sand and gravel particles and analyze the sand and gravel particle crushing process.

9. The quality detection system for crushed sand and gravel particles according to claim 8, characterized in that: The crushing effect evaluation unit sets the crushing standard threshold based on the sand and gravel crushing standard, performs a preliminary comparative evaluation with the uniformity index, analyzes the uniformity of the sand and gravel particles, and determines the adjustment of the crushing process. The specific evaluation contents are as follows; When the uniformity index is greater than the crushing standard threshold, it indicates that the sand and gravel particle size distribution is uneven, and an early warning for crushing process adjustment is generated. When the uniformity index is ≤ the crushing standard threshold, it means that the sand and gravel particle size distribution is uniform and no adjustment is required.

10. The quality detection system for crushed sand and gravel particles according to claim 1, characterized in that: The screen state evaluation module includes a screening rate analysis unit and a screen state evaluation unit; The screening rate analysis unit installs optical sensors at the outlet and inlet of the screen. When the preliminary assessment shows that the sand and gravel particle size distribution is uniform, the optical sensors are activated to count the number of sieved particles and the number of incoming particles in real time, and the number of sieved particles and the number of incoming particles are given to calculate the pass rate of the screen. The screen state assessment unit sets the pass threshold by the user according to the production process requirements, and then performs a secondary comparative assessment with the pass rate to analyze the damage of the screen. The specific assessment content is as follows; When the pass rate is greater than the pass threshold, it indicates that the sieve screening effect is normal and no generation operation is required; When the pass rate is greater than twice the pass threshold, it indicates that the screen screening effect pass rate R is abnormal, and the screen is judged to be damaged. At this time, the first screen warning is generated, prompting you to replace the screen. When the pass rate is ≤ the pass threshold and the time is > 5 seconds, it indicates that the screening effect of the screen is abnormal and the screen surface is judged to be blocked. At this time, a second screen warning is generated to prompt the screen surface to be cleaned.

Images