Research On Audio-visual Fusion Detection Method For Longitudinal Tear Of Conveyor Belt Based On Machine Learning

In mining operation,coal mine transport safety is an important guarantee of coal mine production safety.Belt conveyor is widely used in coal mine material transportation because of its large transport capacity,high stability,easy installation and maintenance and other characteristics.It is an important connection equipment for coal mining and sorting.Conveyor belt is one of the most important part of the belt conveyor in coal mine,in the process of transportation roadway of doping in the raw coal gangue and used for fixed metal bolt and other solid impurities in the production of coal mine could be cut even tear conveyor belt,by comparing the conveyor belt injury accident cases found that the longitudinal tear is the most common,Due to the long transportation distance of the materials in the coal mine,if the longitudinal tearing accident occurs,the whole conveyor belt may be replaced,which will cause the normal production of the coal mine to be unable to recover for a long time,thus causing huge economic losses.But when the coal mine staff fail to detect the accident quickly,it will lead to the blockage of the coal mine roadway,and even cause casualties.Therefore,the detection of conveyor belt longitudinal tear is an important problem to be solved in coal mine safety production.In this paper,audio-visual information representing the damage state of the conveyor belt was introduced into the detection of longitudinal tear of the conveyor belt,and the model and method of audio-visual fusion were studied,so as to break through many limitations of the current detection methods of longitudinal tear of the conveyor belt and provide a new idea for the complex problems in the detection of longitudinal tear of the conveyor belt.In this paper,based on machine learning and combining the complementary characteristics of vision and sound,an audio-visual fusion detection method for longitudinal tearing of conveyor belt based on machine learning was proposed to improve the detection accuracy of longitudinal tearing of conveyor belt in the actual environment of coal mine.The main research content of this method consists of two parts: Firstly,a detection method of conveyor belt longitudinal tear based on traditional machine learning is proposed.Firstly,images and sounds of conveyor belt damage collected by visible CCD and microphone array are processed and analyzed to extract the image and sound features of the normal operation,longitudinal tear and scratch of the conveyor belt respectively.Then,the extracted different features were fused through series,data standardization and PCA dimension reduction.Finally,the fused audio-visual features were input into the machine learning algorithm to construct the audio-visual processing model for longitudinal tearing of the conveyor belt.Second in order to more effective extraction conveyer belt longitudinal tearing sound and image characteristics,puts forward the conveyer belt longitudinal tearing detection method based on the deep learning,will first transport damage sound event into a sonogram,and then in order to obtain the movement of the conveyor belt injury characteristics,build the 3 d audio-visual convolution neural networks,Finally,continuous acoustic spectra and multi-frame conveyor belt damage images in the same time were used as the input layer of the convolutional neural network,and the method of feature extraction and fusion for this task was automatically constructed according to different conditions of conveyor belt damage.Based on the laboratory simulation of the underground working environment of coal mine,this paper verified the detection method in Python compiled environment by means of Open CV library and Tensor Flow framework test.The experimental results show that the method has higher accuracy than the traditional visual detection method in the dark and dusty environment.