Belt longitudinal tear detection method based on neural network

Abstract

The present disclosure relates to a belt longitudinal tear detection method based on a neural network, comprising: S1, constructing a convolutional neural network algorithm model, the convolutional neural network algorithm model comprising a first convolutional network and a second convolutional network; S2, training the convolutional neural network algorithm model through a preset training set, the preset training set comprising a first picture set containing longitudinal tear features and a second picture set containing border annotations for the longitudinal tear features; S3, collecting image information of the coal conveyor belt area; S4, inputting the image information into the convolutional neural network algorithm model; S5, judging whether there is a longitudinal tear feature in the image information, and outputting a picture frame containing border annotations. The belt longitudinal tear detection method based on a neural network provided by the present invention has the beneficial effect of high accuracy compared with general algorithms based on edge detection or pixel recognition.

Description

Technical Field

[0001] The invention relates to a production monitoring method, in particular to a belt longitudinal tearing detection method based on a neural network. Background Art

[0002] Coal conveyor belts are important equipment for fuel transportation in thermal power plants. Coal conveyor belts are generally divided into ordinary belts and steel rope core belts. Thermal power plants usually use ordinary belts, and their bandwidth is generally about 1-1.4m. As the bandwidth increases, the price will become more and more expensive. Due to the high speed of the belt conveyor, the belt may be torn due to equipment or other reasons during transportation. Once it occurs, the tearing range is generally tens of meters or hundreds of meters in length, which may cause economic losses ranging from tens of thousands to hundreds of thousands. At the same time, it may directly affect the normal operation of the unit. Although some units have installed anti-belt tearing devices on the coal conveying system, in actual operation, due to various problems such as system design and signal transmission, the occurrence of belt tearing cannot be effectively prevented. Therefore, it is an important task to fully detect, prevent and deal with the tearing of the coal conveyor belt in a timely manner to ensure the normal production of the unit. Compared with timely handling afterwards, the tearing detection of the coal conveyor belt is a key link in prevention and timely disposal. Generally speaking, belt tearing detection can usually be achieved through manual inspection or video monitoring. With the development of science and technology, there are currently applications for belt tearing detection based on machine vision. For example, Chinese patent application CN105293003A discloses a method for detecting longitudinal tearing of a belt based on machine vision. In this method, the image of the relevant area of ​​the coal conveyor belt is segmented, and then a local pattern is assigned to the segmented area. The pixel pattern is the local pattern whose gray value is closest to the local pattern mean; then the pixels with the geometric feature of line structure are grouped together; further, the 2nd to 4th order Zernike moments of the pixels in the line structure area are calculated, and the template matching method is used to determine whether it is a line. If it matches, it is considered that the belt has longitudinal tearing, otherwise it is normal. However, the inventors found that the above method based on the calculation of pixel gray value has low accuracy and is prone to misidentification. Summary of the invention

[0003] In view of the above problems existing in the prior art, an object of the present invention is to provide a neural network-based belt longitudinal tear detection method with higher detection accuracy.

[0004] In order to achieve the above object, one aspect of the present invention provides a method for detecting longitudinal tearing of a belt based on a neural network, comprising:

[0005] S1. Constructing a convolutional neural network algorithm model, wherein the convolutional neural network algorithm model includes a first convolutional network and a second convolutional network;

[0006] S2. Training the convolutional neural network algorithm model using a preset training set, wherein the preset training set includes a first picture set containing vertical tearing features and a second picture set containing border annotations for the vertical tearing features;

[0007] S3, collecting image information of the coal conveyor belt area;

[0008] S4, inputting the image information into the convolutional neural network algorithm model;

[0009] S5. Determine whether there is a vertical tearing feature in the image information, and output a picture frame including a border annotation.

[0010] Preferably, the first convolutional network includes a first convolutional layer and a pooling layer, the first convolutional layer is configured to perform convolution processing on the first picture set to obtain a first picture subset containing vertical tearing features; the pooling layer is configured to perform pooling operation on the first picture subset to obtain a first feature vector with a dimension of M.

[0011] Preferably, the second convolutional network includes a second convolutional layer and a feature extraction layer, the second convolutional layer is configured to perform convolution processing on the second picture set to obtain a second picture subset including vertical tearing features; the feature extraction layer is configured to extract M second feature vectors corresponding to M regions respectively based on the second picture subset.

[0012] Preferably, the convolutional neural network further includes a combination layer, and the combination layer is configured to combine the first feature vector and the second feature vector to obtain M combination vectors.

[0013] Preferably, the convolutional neural network further comprises a border regression layer, wherein the border regression layer is configured to output a predicted vertical tear area in the second image subset based on the M combined vectors, and the vertical tear area is output with a visual border annotation.

[0014] Preferably, the first convolutional layer and the second convolutional layer share the same convolutional layer.

[0015] Preferably, when performing vector combination, the combination layer performs the combination through a support vector machine.

[0016] Preferably, the border regression layer includes a first hidden layer and a second hidden layer, wherein the first hidden layer is configured to predict the probability of each region of the first picture subset; and the second hidden layer is configured to generate a visual border in the second picture subset.

[0017] Compared with the prior art, the neural network-based belt longitudinal tear detection method provided by the present invention can be trained separately through the determined longitudinal tear pictures and longitudinal tear pictures with frame annotations to obtain the trained convolutional neural network algorithm model, so that the pictures input by the image acquisition device can first pass through the first convolutional network to obtain a subset of pictures that may have longitudinal tearing, and then pass through the second convolutional network for secondary screening, and then through feature vector combination and frame regression, the longitudinal tear pictures containing frame annotations are output to the human-computer interface, which is convenient for staff to view. Compared with the general algorithms based on edge detection or pixel recognition, the present invention has the beneficial effect of high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The present invention is a flowchart of a method for detecting longitudinal tearing of a belt based on a neural network.

[0019] Figure 2 This is a structural block diagram of the system used by the neural network-based belt longitudinal tear detection method of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0021] Various aspects and features of the present invention are described herein with reference to the accompanying drawings.

[0022] These and other characteristics of the invention will become apparent from the following description of a preferred form of embodiment given as a non-limiting example, with reference to the accompanying drawings.

[0023] It should also be understood that, although the invention has been described with reference to some specific examples, those skilled in the art will be able to realize many other equivalent forms of the invention that have the characteristics recited in the claims and are therefore within the scope of protection defined thereby.

[0024] The above and other aspects, features and advantages of the present invention will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings.

[0025] like Figure 1 As shown, a method for detecting longitudinal tearing of a belt based on a neural network is provided in one embodiment of the present invention, comprising:

[0026] S1. Constructing a convolutional neural network algorithm model, wherein the convolutional neural network algorithm model includes a first convolutional network and a second convolutional network;

[0027] S2. Training the convolutional neural network algorithm model using a preset training set, wherein the preset training set includes a first picture set containing vertical tearing features and a second picture set containing border annotations for the vertical tearing features;

[0028] S3, collecting image information of the coal conveyor belt area;

[0029] S4, inputting the image information into the convolutional neural network algorithm model;

[0030] S5. Determine whether there is a vertical tearing feature in the image information, and output a picture frame including a border annotation.

[0031] It is worth noting that the construction of the convolutional neural network algorithm model is an existing technology in the prior art. The difference is that Figure 2 As shown, the first convolutional network in the present invention includes a first convolutional layer and a pooling layer, the first convolutional layer is configured to perform convolution processing on the first picture set to obtain a first picture subset containing vertical tearing features; the pooling layer is configured to perform pooling operation on the first picture subset to obtain a first feature vector with a dimension of M. Similarly, correspondingly, the second convolutional network in the present invention includes a second convolutional layer and a feature extraction layer, the second convolutional layer is configured to perform convolution processing on the second picture set to obtain a second picture subset containing vertical tearing features; the feature extraction layer is configured to extract M second feature vectors corresponding to M regions respectively based on the second picture subset.

[0032] Furthermore, in the present invention, the convolutional neural network also includes a combination layer, the combination layer is configured to combine the first feature vector and the second feature vector to obtain M combination vectors, and the combination layer combines the vectors through a support vector machine. The convolutional neural network also includes a border regression layer, the border regression layer is configured to output a predicted vertical tear area in the second image subset based on the M combination vectors, and the vertical tear area is output with a visual border annotation.

[0033] Moreover, in some improved schemes, in order to further optimize the convolutional neural network algorithm model and improve the computing speed and accuracy, the first convolutional layer and the second convolutional layer share the same convolutional layer.

[0034] Various operations or functions are described herein, which can be implemented or defined as software codes or instructions. Such content can be directly executable ("object" or "executable" form) source code or differential code ("increment" or "patch" code). The software implementation of the embodiments described herein can be provided via a product in which code or instructions are stored or via a method of operating a communication interface to send data via a communication interface. A machine or computer readable storage medium can enable a machine to perform the described functions or operations, and includes any mechanism for storing information in a form accessible by a machine (e.g., a computing device, an electronic system, etc.), such as a recordable / unrecordable medium (e.g., a read-only memory (ROM), a random access memory (RAM), a disk storage medium, an optical storage medium, a flash memory device, etc.). The communication interface includes any mechanism for connecting to any of the hardwired, wireless, optical, etc. media to communicate with another device, such as a memory bus interface, a processor bus interface, an Internet connection, a disk controller, etc. The communication interface can be configured to prepare the communication interface to provide a data signal describing the software content by providing configuration parameters and / or sending a signal. The communication interface can be accessed via one or more commands or signals sent to the communication interface.

[0035] The above embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention. The protection scope of the present invention is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.

Claims

1. A belt longitudinal tear detection method based on a neural network, including: S1. Constructing a convolutional neural network algorithm model, wherein the convolutional neural network algorithm model includes a first convolutional network and a second convolutional network; S2. Training the convolutional neural network algorithm model using a preset training set, wherein the preset training set includes a first picture set containing vertical tearing features and a second picture set containing border annotations for the vertical tearing features; S3, collecting image information of the coal conveyor belt area; S4, inputting the image information into the convolutional neural network algorithm model; S5, determining whether there is a vertical tearing feature in the image information, and outputting a picture frame including a border annotation; The first convolution network includes a first convolution layer and a pooling layer, the first convolution layer is configured to perform convolution processing on the first picture set to obtain a first picture subset containing vertical tearing features; the pooling layer is configured to perform pooling operation on the first picture subset to obtain a first feature vector with a dimension of M, the second convolution network includes a second convolution layer and a feature extraction layer, the second convolution layer is configured to perform convolution processing on the second picture set to obtain a second picture subset containing vertical tearing features; the feature extraction layer is configured to extract M second feature vectors corresponding to M regions respectively based on the second picture subset, the convolution neural network also includes a combination layer, the combination layer is configured to combine the first feature vector and the second feature vector to obtain M combination vectors; the convolution neural network also includes a border regression layer, the border regression layer is configured to output a predicted vertical tearing region in the second picture subset based on the M combination vectors, and the vertical tearing region is output with a visual border annotation; the border regression layer includes a first hidden layer and a second hidden layer, the first hidden layer is configured to predict the probability of each region of the picture of the first picture subset; the second hidden layer is configured to generate a visual border in the second picture subset; The first convolutional layer and the second convolutional layer share the same convolutional layer.

2. According to the method of claim 1, when the combination layer performs vector combination, the combination is performed through a support vector machine.

Images