Damage positioning method and system for belt flaw detection

Abstract

The invention relates to a damage positioning method and system for belt flaw detection, and belongs to the field of image data processing and visual identification. According to the invention, through a deep neural network algorithm, efficient and accurate identification of belt surface damage is realized; the edge gnawing detection method provided by the invention saves computing resources; highly automatic damage detection and positioning are realized, and manual intervention and personal errors are reduced; a line-scan digital camera, a speed sensor and a specific marker are combined to accurately position the damage position, so that the mapping of the damage position is more reliable, and maintenance personnel can conveniently, timely and accurately carry out maintenance; the system can stably work under complex environmental conditions, solves the problem of positioning failure caused by factors such as number missing and poor illumination, and improves the detection reliability; compared with a serial number-based damage positioning method, the damage positioning method provided by the invention can provide a specific damage position according to the reference object around the belt, so that the situation that the maintenance personnel cannot find the damage position after shutdown is avoided.

Description

Technical Field

[0001] The invention belongs to the field of image data processing and visual recognition, and relates to a damage locating method and system for belt flaw detection. Background Art

[0002] At present, the main damage detection methods for belt conveyor belts include visual detection, ultrasonic detection, electromagnetic wave detection, infrared thermal imaging detection, vibration and resistivity measurement, etc. Visual detection technology uses high-definition cameras and image processing technology to monitor the damage, wear and cracks on the belt surface in real time; ultrasonic detection uses ultrasonic sensors to detect cracks and damage inside the belt, which is suitable for detecting the deep structure of the belt; electromagnetic wave detection uses electromagnetic wave sensors to detect abnormalities on the surface and inside of the belt, which can achieve non-contact detection; infrared thermal imaging uses thermal imaging technology to detect the temperature change of the belt during operation, which is used to detect abnormal conditions such as friction and overheating; vibration monitoring installs vibration sensors to monitor the vibration of the belt during conveying, which can detect abnormal vibration and imbalance; resistivity measurement measures the resistivity change of the belt material to evaluate its health status, which is usually used for older belt detection.

[0003] The above-mentioned visual inspection technology has gradually become the mainstream method of belt flaw detection with the development of artificial intelligence technology. At the same time, due to the mature application of linear array camera technology, the visual flaw detection technology based on line scan images has been well applied; the linear array camera has a high resolution and can capture subtle defects and details on the belt surface, such as cracks, wear and scratches, which makes the detection more accurate and reliable; the linear array camera can perform continuous scanning at high speed, which is suitable for detecting high-speed running belts, which is very important for real-time monitoring and timely detection of problems, especially in occasions that require continuous and large-area detection; however, the linear array camera has a relatively strict requirement for uniform lighting, so it is necessary to add a strong fill light device to the detection surface; at the same time, the linear array camera needs to match the motion process during the shooting process, so it is necessary to add a position feedback link, that is, a speed sensor. Therefore, this belt flaw detection system based on linear array cameras has higher practical value than traditional linear array cameras, is a method worthy of recommendation in the current belt flaw detection field, and is also the basic system relied on by the present invention.

[0004] The currently more popular damage location method of the belt flaw detection system based on machine vision technology is also based on visual technology. The principle is: there are many ways to connect conveyor belts on the market, usually metal buckle joints, cold glue joints, hot vulcanized joints, etc. The purpose of the joint is to connect multiple conveyor belts into a ring to increase the distance of the belt. The joint number is usually engraved at each joint, such as "①, ②", and each number corresponds to a joint. Then, the number detection can be performed through machine learning methods, and the position of the number on the conveyor belt can be recorded through visual detection technology; after detecting the damage on the conveyor belt, the position of the joint number near it is used as a reference to obtain the damage position between two numbers of the belt or the distance from a certain number.

[0005] Although the positioning method based on vision is also a relatively new and popular technology in recent years, its principle of identifying the number at the belt joint is similar to the license plate number recognition, which is already a relatively mature application method. However, its application in the detection of joint numbers of industrial conveyor belts faces many technical defects. For example, defect 1: the damage detection algorithm based on the deep learning network model is used. Since its reasoning performance is greatly affected by hardware resources, it needs to occupy more computing resources when applied on site, which will greatly increase the construction cost of the system. Defect 2: The prerequisite for accurately identifying the joint number is that the belt joint must be numbered. However, the joint of the belt may not be engraved with numbers during production, which makes the method of obtaining the belt damage position based on numbers invalid; similarly, if the number is on the non-bearing surface, since the linear array camera needs to be installed on the bearing surface, the task of number recognition cannot be completed at this time; moreover, the goods transported by the transport belt are usually more complicated, such as coal mines, gravel and other dusty items, and the number recognition failure problem will also be caused by factors such as number obstruction and poor lighting. Defect 3: Since the distance between each joint number usually ranges from one hundred meters to several hundred meters, even if it is known which two numbers the damage is between or how many meters away from a certain number, more specific and effective location information cannot be given. Defect 4: If the belt is in operation, the measured defect location information needs to be cached in the system. At this time, in addition to continuing to identify the damage, the system must also identify different numbers, which increases the complexity of the system and reduces the detection efficiency. Defect 5: When the belt is shut down, even if it is known which numbers the damage location is adjacent to, the belt maintenance personnel cannot know the specific location of the number on the belt, and thus cannot perform maintenance work on the damage. From the above defects, it can be concluded that the practicality of the method based on vision to locate belt damage is low. Summary of the invention

[0006] In view of this, the technical problems to be solved by the present invention are: the damage detection model based on deep learning occupies a large amount of computing resources; the visual positioning method is greatly affected by environmental factors and has low detection efficiency; maintenance personnel are unable to quickly know the specific damage points, and its practicality is low.

[0007] In order to solve the above technical problems, the present invention proposes a damage location system and method for belt flaw detection, which detects damage to the belt, such as scratches, potholes, and peeling, through a deep neural network; in order to save computing resources, a belt edge gnawing detection method is proposed on this basis for selection in practical applications, which can achieve better detection effects while reducing computing resources; further, a damage location method is proposed, which calculates the relative distance between the damage and the installation position of the linear array camera when the belt is stopped, combined with specific reference objects around the belt, to obtain the specific location of the damage. This positioning method has high practicality and robustness, and greatly improves the efficiency and practicality of belt maintenance work.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A damage location method for belt flaw detection comprises the following steps:

[0010] Assume that the total length of the belt is L, and the belt located above the horizontal line of the center of the roller is the upper belt, and its length is L / 2; the belt located below the horizontal line of the center of the roller is the lower belt, and its length is L / 2; the roller on the side closer to the linear array camera along the running direction of the belt is the head, and the distance between the linear array camera and the vertical tangent of the outer edge of the head is m; the distance between the reference objects fixed and continuously set around the belt is n, and the reference objects are numbered with the reference object at the head as the starting point;

[0011] When issuing a flaw detection task, the point on the belt corresponding to the linear array camera is set to A;

[0012] When the belt is running, the distance A runs at the current moment is s;

[0013] When the damage point x is identified, the distance A has traveled at the current moment is recorded as s x ;

[0014] When the belt is locked, the distance traveled by A at the current moment is s. end , the point on the belt corresponding to the position of the linear array camera is set to B; calculate the distance s of the damage point x relative to B along the running direction of the belt Bx , expressed as:

[0015] s Bx =(s end -s x )%L

[0016] Among them, send Indicates the distance the belt travels from the time the task is issued to the time the belt is locked; s x Indicates the distance the belt has traveled when the damage point is identified; % indicates the remainder operation, L indicates the total length of the belt; s end -s x It indicates the distance that the damage point x moves when the belt is locked;

[0017] Use the distance s between the damage point x and B along the belt running direction Bx As well as the distance n between the reference objects and the reference object number, the position of the damage point x relative to the specific reference object is determined.

[0018] Furthermore, the distance s between the damage point x and B along the belt running direction is used. Bx As well as the distance n between the reference objects and the reference object number, the position of the damage point x relative to the specific reference object is determined, specifically:

[0019] If Bx ≤(L / 2-m), then the damage point x is located in the lower belt, s away from the machine head Bx +m, the reference number code_H near the damage point x is:

[0020]

[0021] If Bx >(Lm), then the position of the damage point x is in the lower belt, and the distance from the machine head is s Bx -(Lm), the reference object number code_H near the damage point x is:

[0022]

[0023] If (L / 2-m)<s Bx <(Lm), then the position of the damage point x is on the upper belt, and the distance from the machine head is Ls Bx -m, the reference number code_H near the damage point x is:

[0024]

[0025] in, Indicates rounding down.

[0026] Further, the method for identifying the damage point is to detect and identify the scratches, edge gnawing, potholes and peeling damage of the belt by using a convolutional neural network; and to detect and identify the edge gnawing damage of the belt by using a non-convolutional neural network, including the following steps:

[0027] The gray value difference between the belt and the environment in the image is obtained by image processing technology to extract the belt edge;

[0028] Perform dilation operation on the extracted edge image to fill the pixels near the edge to make the edge more obvious;

[0029] Perform a difference operation on the expanded edge image and the current edge image to obtain a difference image;

[0030] Perform connected domain analysis on the difference image, find all connected domains, and calculate the radius of the inscribed circle of each connected domain;

[0031] Customize the threshold value and compare the radius of the inscribed circle with the threshold value to determine the edge damage point.

[0032] Furthermore, the step of obtaining the gray value difference between the belt and the environment in the image by image processing technology to extract the belt edge is specifically as follows:

[0033] Assume that the input image is I(x,y) and its grayscale value is G(x,y). The edge information in the input image I(x,y) is extracted by edge detection algorithm, which is expressed as:

[0034] E(x,y)=ED(G(x,y))

[0035] Where x represents the horizontal position of the pixel; y represents the vertical position of the pixel; E(x,y) represents the binary image after edge detection, the edge part of the binary image is marked as white, and the non-edge part is marked as black; ED represents the edge detection algorithm;

[0036] The step of performing a dilation operation on the extracted edge image to fill pixels near the edge to make the edge portion more obvious is specifically:

[0037] The expansion operation is expressed as:

[0038] D(x,y)=D(E(x,y),k)

[0039] Where D represents the dilation operation, k represents the structural element of the dilation operation, which makes the white area (foreground) in the image more connected by expanding it; D(x, y) represents the binary image after dilation;

[0040] The step of performing a difference operation on the expanded edge image and the current edge image to obtain a difference image is specifically as follows:

[0041] The difference image is expressed as:

[0042] Δ(x,y)=D(x,y)-E(x,y)

[0043] Among them, Δ(x,y) represents the difference image, which indicates the difference between the dilated edge and the current edge;

[0044] The steps of performing connected domain analysis on the difference image, finding all connected domains, and calculating the radius of the inscribed circle of each connected domain are specifically as follows:

[0045] Assumption C i represents the i-th connected domain, then the connected domain C i The radius of the inscribed circle R(C i )for:

[0046] R(C i )=max (x,y)∈Ci (min(dist((x,y),boundary(C i ))))(4)

[0047] Among them, dist((x,y),boundary(C i )) represents the distance from the point (x, y) to the boundary of the connected domain;

[0048] The step of customizing the threshold and comparing the radius of the inscribed circle with the threshold to determine the edge-biting damage point is specifically as follows:

[0049] The threshold value is set to T. If the radius of the inscribed circle R(C i ) is greater than the threshold value T, then the radius of the inscribed circle R(C i ) is the edge damage point. If the radius of the inscribed circle R(C i ) is less than the threshold, then the inscribed circle radius R(C i ) is the non-edge-biting damage point.

[0050] Furthermore, the linear array camera has a distance measurement function, and the s, s end 、s x Acquired by the line array camera.

[0051] Further, the method for realizing the ranging function is:

[0052] The linear array camera records a timestamp when taking each frame of image;

[0053] The speed sensor is used to obtain the moving speed of the belt, and the distance the belt moves between two adjacent frames of images is calculated by combining the timestamp and speed information.

[0054] The distances the belt moves between two adjacent frames of images are accumulated to obtain the total distance the belt moves from the start of shooting to any moment.

[0055] A damage location system for belt flaw detection, comprising:

[0056] Linear array camera, which collects images of the belt surface and transmits them to the server;

[0057] Fill light, which provides uniform lighting for the line scan camera under various lighting conditions;

[0058] Speed ​​sensor, which monitors the belt running speed and feeds it back to the linear array camera, coordinating image acquisition and belt running synchronization;

[0059] Servers, including:

[0060] The image processing module pre-processes the images collected by the linear array camera to improve the image clarity; extracts the feature information in the image and transmits it to the damage location module;

[0061] The damage location module uses the feature information in the image to identify the damage points on the belt in real time and transmits them to the data storage and analysis module;

[0062] The data storage and analysis module uses the image data and the damage point to determine the damage point type and locate the damage point position, and stores the damage point type, the damage point position and the image data.

[0063] Furthermore, it also includes: an alarm display host computer, which is connected to the data storage and analysis module to display the image data, the type of damage point, the location of the damage point and the time when the damage point was generated, and supports human-computer interaction.

[0064] Furthermore, the extracting feature information from the image specifically includes: extracting edge information from the image using an edge detection algorithm.

[0065] Furthermore, the data storage and analysis module only stores the types of damage points and their specific location information detected within the last two turns of the belt at the current moment.

[0066] The beneficial effects of the present invention are:

[0067] Efficient detection: Through the deep neural network algorithm, efficient and accurate identification of belt surface damage is achieved, greatly improving the detection efficiency; at the same time, the proposed edge detection method is efficient and saves computing resources; the whole system realizes highly automated damage detection and positioning, reducing manual intervention and human errors. Accurate positioning: Combined with linear array cameras, speed sensors and specific markers, the location of the damage can be accurately located, making the damage location mapping more reliable and facilitating maintenance personnel to carry out maintenance work in a timely and accurate manner. Strong robustness: The system can work stably under complex environmental conditions, overcomes the problem of positioning failure caused by missing numbers, poor lighting and other factors, and improves the reliability of detection. High practicality: Compared with the damage location method based on numbers, the method of the present invention can provide a specific damage location, avoiding the maintenance personnel from being unable to find the damage location after shutdown; at the same time, it is not limited by the belt material, width and speed, and has broad application prospects.

[0068] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0070] Figure 1 This is a schematic diagram of the installation of a damage location system for belt flaw detection according to an embodiment of the present invention;

[0071] Figure 2 It is a flow chart of a belt edge damage detection method according to an embodiment of the present invention;

[0072] Figure 3 This is a schematic diagram of the belt structure of an embodiment of the present invention;

[0073] Figure 4 The present invention is a flowchart of a method for locating damage in belt flaw detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0074] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0075] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0076] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0077] The damage location system for belt flaw detection proposed in the embodiment of the present invention mainly includes: a linear array camera, which collects images of the belt surface and transmits them to a server; a fill light, which provides uniform lighting for the linear array camera under various lighting conditions; a speed sensor, which monitors the running speed of the belt and feeds back to the linear array camera, and coordinates image acquisition and belt operation synchronization; a server, including: an image processing module, which preprocesses the images collected by the linear array camera, such as preprocessing for filtering, denoising and enhancement to improve the clarity of the image; extracts feature information in the image and transmits it to a damage location module; a damage location module, which uses the feature information in the image to identify damage points on the belt in real time and transmits it to a data storage and analysis module; a data storage and analysis module, which stores image data, and uses the image data and damage points to perform image analysis and location to obtain the position of the damage point.

[0078] See also Figure 1, is a schematic diagram of the installation of the damage location system for belt flaw detection according to an embodiment of the present invention; the linear array camera and the fill light are installed above the belt running bearing surface. Since the linear array camera has a relatively high requirement for light, the fill light needs to be kept on when the linear array camera is working; the best installation position of the linear array camera and the fill light is above one side of the belt roller, where there can be no interference from goods all year round, the belt surface is relatively flat, and the detection interference of the belt edge (due to belt deviation, friction between the belt and the roller or the frame, etc., resulting in belt edge wear or damage) is also small; the speed sensor is installed above the non-bearing surface of the belt under the belt; the server and other equipment can be installed in the equipment chamber or the machine room, and the linear array camera can be connected to the server and the speed sensor by a network cable, and the power supply and control of the linear array camera can be a four-core cable. When the belt is running, the speed sensor is driven to rotate, and the speed sensor feeds back the pulse signal to the linear array camera, and the linear array camera transmits the collected image to the server through the image processing module and the damage location module for damage detection; the linear array camera adopts the frame trigger mode and starts shooting after receiving the external signal. The user can customize the line height (pixel height) of the captured image, and the line array camera will continue to shoot at the preset line frequency (number of lines shot per second) until the specified line height is reached. This method enables the line array camera to capture images of the continuously moving belt surface. Since the belt is in continuous motion, and the line array camera continues to shoot at the set line frequency after the frame is triggered, the movement distance of the belt can be determined by analyzing the captured image sequence, that is, it has a distance measurement function. Specifically, distance measurement can be achieved through the following methods:

[0079] 1. Timestamp and speed calculation:

[0080] Line scan cameras record a timestamp when capturing each frame.

[0081] The speed of the belt is obtained by the speed sensor. Combining the timestamp and speed information, the distance the belt moves between two adjacent frames of images can be calculated.

[0082] By adding up these distances, the total distance the belt has traveled from the start of filming to any given moment can be obtained.

[0083] 2. Image feature matching:

[0084] In the captured image sequence, look for frames with specific features, such as splice numbers, marks or defects on the belt.

[0085] By matching the positions of these feature frames in the image sequence, the relative distances between the features can be calculated.

[0086] If the distance between the characteristic frames is known or can be measured by other means, the distance the entire belt moves can be calculated from the proportional relationship.

[0087] 3. Direct measurement:

[0088] If the parameters such as the installation position, angle and focal length of the line scan camera are known, and the texture or markings on the belt surface are clear enough, the actual belt movement distance can be directly inferred by measuring the pixel distance in the image.

[0089] This method requires higher image resolution and accurate calibration of line array camera parameters.

[0090] When the server identifies damage through the damage detection algorithm, it will record the current detected distance; when the belt stops, the current detected distance will also be recorded; the installation position of the integrated linear array camera can calculate the relative position of the damage when the belt stops, and then map the position to the position of specific markers around the belt, which can give a reference position of the damage, facilitating maintenance personnel to carry out timely inspections.

[0091] There are many types of common belt damage, such as scratches, edge dents, potholes, peeling, etc. Different types of damage require different detection methods. If all of them are detected and identified through convolutional neural networks, the computing resources occupied will be relatively large, which is not suitable for analysis and application at the edge. This embodiment proposes a non-convolutional neural network belt edge dent detection method, which can achieve better detection results while reducing computing resources. Please refer to Figure 2 , is a flow chart of a belt edge gnawing damage detection method according to an embodiment of the present invention; the belt edge gnawing damage detection method is specifically implemented as follows:

[0092] 1. Belt edge extraction:

[0093] First, the grayscale value difference between the belt and the environment in the image is obtained through image processing technology to extract the belt edge. Assume that the input image is I(x,y), and its grayscale value is G(x,y); x represents the position of the pixel in the horizontal direction (usually called the column). y represents the position of the pixel in the vertical direction (usually called the row); through edge detection algorithms, such as Sobel Operator and Canny Edge Detection, to extract the edge information in the input image I(x,y):

[0094] E(x,y)=ED(G(x,y)) (1)

[0095] Wherein, ED represents the edge detection algorithm, and E(x, y) is the binary image after edge detection. In this binary image, the edge part is marked as 1 (or white), and the non-edge part is marked as 0 (or black).

[0096] 2. Edge pixel filling:

[0097] Perform a dilation operation on the extracted edge image E(x,y) to fill the pixels near the edge to make the edge more obvious. The formula for the dilation operation algorithm is as follows:

[0098] D(x,y)=D(E(x,y),k) (2)

[0099] Where D represents the dilation operation, k is the structural element of the dilation operation, which expands the white area (foreground) in the image to make the white area more connected; its specific implementation is to convolve a structural element with each pixel in the image. If the structural element covers the foreground pixel, the pixel is set to the foreground (i.e. white), otherwise it is set to the background (i.e. black); D(x, y) is the binary image after dilation, and the structural element is usually a small matrix of size n×n.

[0100] 3. Difference image calculation:

[0101] Perform a difference operation on the expanded edge image D(x, y) and the current edge image E(x, y) to obtain the difference image:

[0102] Δ(x,y)=D(x,y)-E(x,y) (3)

[0103] Among them, Δ(x, y) is the difference image, which represents the difference between the dilated edge and the current edge.

[0104] 4. Damage radius detection:

[0105] Perform connected domain analysis on the difference image, find all connected domains, and calculate the radius of the inscribed circle of each connected domain. Assume C i represents the i-th connected domain, then the connected domain C i The radius of the inscribed circle R(C i )for:

[0106] R(C i )=max (x,y)∈Ci (min(dist((x,y),boundary(C i ))))(4)

[0107] Among them, dist((x,y),boundary(C i )) represents the distance from the point (x, y) to the boundary of the connected domain.

[0108] 5. Judgment of gnawing edge:

[0109] The threshold value is set to T, and the radius of the inscribed circle R (C i) is greater than the threshold. If it is greater than the threshold, it is determined to be edge gnawing. If it is less than the threshold, it is determined to be no edge gnawing. What can be customized here is that after the connected domain analysis, a gap in the belt will have multiple inscribed circles. Therefore, it can be determined that the edge gnawing is only determined when the radius of one or more of the inscribed circles exceeds the threshold.

[0110] This embodiment also proposes a belt damage location method: when the belt is running, after the server recognizes the damage, it records the current detection distance, and records it once for each damage recognized; when the belt stops, the detection distance at this time is recorded. Then, the relative position of the damage when the belt stops can be calculated based on the installation position of the integrated linear array camera, and then the position is mapped to the position of specific markers around the belt, which can give a reference position of the damage, facilitating maintenance personnel to carry out timely inspections. The specific implementation of this belt damage location method must meet the following prerequisite requirements:

[0111] 1. Get the total length of the belt;

[0112] 2. Obtain the overall structure of the belt, such as the number of drive cylinders and steering cylinders and the distance between the drive cylinders and steering cylinders;

[0113] The data required for the above two steps are usually described in detail when the belt system is first built, so it is easy to obtain; it can also be measured again to verify or update the above data.

[0114] Since the construction of the belt system is a prerequisite for damage location, there are slight differences in the positioning calculation methods under different belt structures. Therefore, this embodiment uses one of the belt structures to illustrate the method. The belt is supported by continuously arranged H frames, each H frame has a different number, the upper belt and the lower belt of the belt share the same H frame number, and the distance between adjacent H frames is constant; in industrial applications, due to the heavy weight of the goods transported by the belt, the H frames required are relatively dense, and the interval between adjacent H frames is generally about 1 meter. If the H frame is used as a specific marker around the belt, maintenance personnel can quickly find the damage point near the location of the H frame through the H frame number. Similarly, fixed reference objects other than the H frame can also be determined as suitable markers based on on-site applications to help maintenance personnel find damage points nearby through the markers. Please refer to Figure 3 , is a schematic diagram of the belt structure of an embodiment of the present invention; assuming that the belt is composed of a head roller and a tail roller, and the overall structure is relatively standard, approximately the upper belt and the lower belt are equidistant, the upper belt: the belt located above the roller; the lower belt: the belt located below the roller; assuming that the linear array camera is installed on the lower belt, the belt runs counterclockwise, that is, Figure 3 Shown is a frontal view with the nose on the left and the tail on the right.

[0115] It is known that the total length of the belt is L, and the linear array camera is deployed at a distance m from the head of the belt. Taking the standard belt structure as an example, the upper and lower belts of the belt are equidistant and the length is L / 2. After receiving the inspection task issued, the system starts to perform belt flaw detection. The key positions and distances during the operation of the belt are analyzed as follows:

[0116] 1. When issuing a flaw detection task, the position of the linear array camera corresponds to point A on the belt.

[0117] 2. When the belt is running, record the distance s that point A has traveled at the current moment; s is a variable; the value of s is the distance that point A has traveled, which can be obtained through a linear array camera with a distance measurement function.

[0118] 3. When the belt is locked, the current running distance is recorded as s end , the position of the linear array camera corresponds to point B on the belt.

[0119] 4. When the damage point x is identified, the distance traveled by point A at the current moment is recorded as s x ;

[0120] 5. Assume the distance n between adjacent H frames, assume that H frame No. 0 is at the head of the belt conveyor, and number the H frames according to their relative positions.

[0121] In this method, the calculation of damage location information strictly follows the conditions of belt locking. Specifically, this means that after the detection task is started, all potential damage location information detected by the system must be finally determined and calculated at the moment the belt is locked. This process ensures the accuracy of damage location and emphasizes that belt locking is a key prerequisite for calculating damage location information. In other words, only when the belt stops running and is in a locked state, the system will formally calculate the specific location information of the damage based on the collected detection data. Because only when the belt is locked does the damage location information have practical significance and application value. Please refer to Figure 4 , is a flow chart of a damage location method for belt flaw detection according to an embodiment of the present invention. Therefore, after the detection task is issued, when the belt is locked, the method steps for calculating the location of the damage point are as follows:

[0122] 1. Calculate the distance of the damage point x relative to the linear array camera in the positive direction (along the running direction of the belt):

[0123] When the belt is locked, the positive distance of the damage point x relative to the point where the linear array camera is located (point B) is:

[0124] s Bx =(s end -s x )%L (5)

[0125] Among them, s endIndicates the distance the belt travels from the time the task is issued to the time the belt is locked; s x = represents the distance the belt has traveled when the damage point is identified; % represents the remainder operation, and L represents the total length of the belt. This formula describes that when the damage x is identified, point A has moved s x m, and the damage point starts to move at this time. When the belt stops, the damage point x moves s end -s x Meters; the total length of the belt is L, which is the distance of one circle, so the modulus is required to calculate the relative distance between the damage point and the linear array camera.

[0126] 2. Determine the specific location of the damage point x:

[0127] Because the reference markers are the H-frames with numbers along the belt, the final calculation result of the system is output in the form of the closest H-frame, and considering the upper and lower belts of the belt, the interval between the H-frames in this embodiment is 1.2 meters. The method for determining the specific position of the damage point x is:

[0128] 1) If s Bx ≤(L / 2-m), then the damage x is located in the lower belt, s away from the machine head Bx +m meters, the nearby H frame number is Indicates rounding down.

[0129] The output positioning information is: "Near the H rack with code_H!".

[0130] 2) If s Bx >(Lm), then the damage x is located in the lower band, s away from the nose Bx -(Lm) meters, the nearby H frame number is Indicates rounding down.

[0131] The output positioning information is: "Near the H rack with code_H!".

[0132] 3) If (L / 2-m)<s Bx <(Lm), then the position of the damage x is on the upper belt, at a distance of Ls from the machine head Bx -m meters, the nearby H frame number is Indicates rounding down.

[0133] The output positioning information is: "Near the H rack with code_H!".

[0134] The parameters involved in the above steps are described as follows:

[0135] L is the total length of the belt, with a custom setting interface reserved;

[0136] m is the distance from the installation position of the linear array camera to the belt length of the machine head, and a custom setting interface is reserved;

[0137] s Bx is the positive distance from the damage point x to the linear array camera position when the belt is locked, and the remainder is always less than the belt length L;

[0138] code_H is the H frame number near the damaged point when the belt stops, and maintenance personnel can find the damaged location based on it.

[0139] In summary, the process of the damage location method of the belt flaw detection system can be described as Figure 4 , when the preconditions are accurate, the accuracy of this method can be controlled within 1.5 meters, and it can be implemented only at the software level, without being affected by environmental factors. It is suitable for belt damage positioning detection in harsh environments such as coal flow transportation systems in coal mines.

[0140] Optionally, to save system cache resources, cache logic can be added as follows: After the task is issued, the system will cache the current detection distance every time it detects a damage. If the belt runs continuously, it will cause a waste of resources. Therefore, a logic that only caches the last two circles can be added, which is explained as follows: the system clears the information cache of the previous two circles every two circles, and the number of circles is framed by distance. For example, 5000 meters of detection is 1 circle. This is equivalent to the normal push of the damage alarm information detected by the system before the belt stops. When the belt stops, the location information of the last two circles of damage is calculated according to the cache and positioning method, and the alarm of the last two circles can be pushed again.

[0141] In summary, in order to solve the relevant technical problems, the present invention provides a belt flaw detection and positioning system, including an analysis server, a linear array camera, a fill light, a speed sensor, and an alarm display host computer software. The alarm event information and processing results of the entire system are uniformly displayed through the host computer software, and the displayed information includes: damage alarm pictures, alarm types, alarm time and location information, etc. The linear array camera is installed above the belt bearing surface near the roller, the fill light is installed in the belt line of sight and illuminates the belt surface; the speed sensor is installed between the upper belt and the lower belt, specifically on the non-bearing surface of the lower belt; the server is installed in the electromechanical chamber or the machine room.

[0142] The belt damage detection algorithm and damage location method described in this invention are both software logics, built into the analysis server. After the entire system is implemented, when the belt is running and a flaw detection task is issued, the system will detect the type of damage in real time and push an alarm picture; when the belt stops, the system will calculate the position according to the positioning method proposed in this article, and push the alarm information of the set time period, so that maintenance personnel can check the belt damage status and damage location, make maintenance decisions in time, and avoid accidental stop of the conveying task, resulting in production suspension or safety accidents.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A damage location method for belt flaw detection, characterized in that: The following steps are involved: Assume that the total length of the belt is L, and the belt located above the horizontal line of the center of the roller is the upper belt, and its length is L / 2; the belt located below the horizontal line of the center of the roller is the lower belt, and its length is L / 2; the roller on the side closer to the linear array camera along the running direction of the belt is the head, and the distance between the linear array camera and the vertical tangent of the outer edge of the head is m; the distance between the reference objects fixed and continuously set around the belt is n, and the reference objects are numbered with the reference object at the head as the starting point; When issuing a flaw detection task, the point on the belt corresponding to the linear array camera is set to A; When the belt is running, the distance traveled by A at the current moment is s; When the damage point x is identified, the distance A has traveled at the current moment is recorded as s x ; When the belt is locked, the distance traveled by A at the current moment is s. end , the point on the belt corresponding to the position of the linear array camera is set to B; calculate the distance s of the damage point x relative to B along the running direction of the belt Bx , expressed as: s Bx =(s end -s x )%L Among them, s end Indicates the distance the belt travels from the time the task is issued to the time the belt is locked; s x Indicates the distance the belt has traveled when the damage point is identified; % indicates the remainder operation, L indicates the total length of the belt; s end -s x It indicates the distance that the damage point x moves when the belt is locked; Use the distance s between the damage point x and B along the belt running direction Bx As well as the distance n between the reference objects and the reference object number, the position of the damage point x relative to the specific reference object is determined.

2. The damage location method for belt flaw detection according to claim 1, characterized in that: The distance s between the damage point x and B along the belt running direction is used Bx As well as the distance n between the reference objects and the reference object number, the position of the damage point x relative to the specific reference object is determined, specifically: If Bx ≤(L / 2-m), then the damage point x is located in the lower belt, s away from the machine head Bx +m, the reference number code_H near the damage point x is: If Bx >(Lm), then the position of the damage point x is in the lower belt, and the distance from the machine head is s Bx -(Lm), the reference object number code_H near the damage point x is: If (L / 2-m)<s Bx <(Lm), then the position of the damage point x is on the upper belt, and the distance from the machine head is Ls Bx -m, the reference number code_H near the damage point x is: in, Indicates rounding down.

3. The damage location method for belt flaw detection according to claim 2, characterized in that: The method for identifying the damage point is to detect and identify the scratches, edge gnawing, potholes and peeling damage of the belt by using a convolutional neural network; and to detect and identify the edge gnawing damage of the belt by using a non-convolutional neural network, including the following steps: The gray value difference between the belt and the environment in the image is obtained by image processing technology to extract the belt edge; Perform dilation operation on the extracted edge image to fill the pixels near the edge to make the edge more obvious; Perform a difference operation on the expanded edge image and the current edge image to obtain a difference image; Perform connected domain analysis on the difference image, find all connected domains, and calculate the radius of the inscribed circle of each connected domain; Customize the threshold value and compare the radius of the inscribed circle with the threshold value to determine the edge damage point.

4. The damage location method for belt flaw detection according to claim 3, characterized in that: The step of obtaining the gray value difference between the belt and the environment in the image by image processing technology to extract the belt edge is specifically as follows: Assume that the input image is I(x,y) and its gray value is G(x,y). The edge information in the input image I(x,y) is extracted by edge detection algorithm, which is expressed as: E(x,y)=ED(G(x,y)) Where x represents the horizontal position of the pixel; y represents the vertical position of the pixel; E(x,y) represents the binary image after edge detection, the edge part of the binary image is marked as white, and the non-edge part is marked as black; ED represents the edge detection algorithm; The step of performing a dilation operation on the extracted edge image to fill pixels near the edge to make the edge portion more obvious is specifically: The expansion operation is expressed as: D(x,y)=D(E(x,y),k) Where D represents the dilation operation, k represents the structural element of the dilation operation, which makes the white area (foreground) in the image more connected by expanding it; D(x, y) represents the binary image after dilation; The step of performing a difference operation on the expanded edge image and the current edge image to obtain a difference image is specifically as follows: The difference image is expressed as: Δ(x,y)=D(x,y)-E(x,y) Among them, Δ(x,y) represents the difference image, which indicates the difference between the dilated edge and the current edge; The steps of performing connected domain analysis on the difference image, finding all connected domains, and calculating the radius of the inscribed circle of each connected domain are specifically as follows: Assumption C i represents the i-th connected domain, then the connected domain C i The radius of the inscribed circle R(C i )for: R(C i )=max (x,y)∈Ci (min(dist((x,y),boundary(C i ))))(4) Among them, dist((x,y),boundary(C i )) represents the distance from the point (x, y) to the boundary of the connected domain; The step of custom setting a threshold value and comparing the inscribed circle radius with the threshold value to determine the edge biting damage point is specifically as follows: the custom setting threshold value is T, if the inscribed circle radius R (C i ) is greater than the threshold value T, then the radius of the inscribed circle R(C i ) is the edge damage point. If the radius of the inscribed circle R(C i ) is less than the threshold, then the inscribed circle radius R(C i ) is the non-edge-biting damage point.

5. The damage location method for belt flaw detection according to claim 2, characterized in that: The linear array camera has a distance measurement function, and the s, s end 、s x Acquired by the line array camera.

6. The damage location method for belt flaw detection according to claim 3, characterized in that: The method for realizing the ranging function is: The linear array camera records a timestamp when taking each frame of image; The speed sensor is used to obtain the moving speed of the belt, and the distance the belt moves between two adjacent frames of images is calculated by combining the timestamp and speed information. The distances the belt moves between two adjacent frames of images are accumulated to obtain the total distance the belt moves from the start of shooting to any moment.

7. A damage location system for belt flaw detection, characterized in that: include: Linear array camera, which collects images of the belt surface and transmits them to the server; Fill light, which provides uniform lighting for the line scan camera under various lighting conditions; Speed ​​sensor, which monitors the belt running speed and feeds it back to the linear array camera, coordinating image acquisition and belt running synchronization; Servers, including: The image processing module pre-processes the images collected by the linear array camera to improve the image clarity; extracts the feature information in the image and transmits it to the damage location module; The damage location module uses the feature information in the image to identify the damage points on the belt in real time and transmits them to the data storage and analysis module; The data storage and analysis module uses the image data and the damage point to determine the damage point type and locate the damage point position, and stores the damage point type, the damage point position and the image data.

8. The damage location system for belt flaw detection according to claim 7, characterized in that: Also includes: The alarm display host computer is connected to the data storage and analysis module to display the image data, the damage point type, the location of the damage point and the time when the damage point was generated, and supports human-computer interaction.

9. The damage location system for belt flaw detection according to claim 8, characterized in that: The extracting feature information from the image specifically includes: extracting edge information from the image using an edge detection algorithm.

10. The damage location system for belt flaw detection according to claim 9, characterized in that: The data storage and analysis module only stores the types of damage points and their specific location information detected within the last two turns of the belt at the current moment.

Images