Research On Image Recognition Of Foreign Bodies In The Process Of Coal Mine Belt Transportation In Complex Environment

As the largest coal producer and consumer in the world,China takes coal as the main energy in the case of limited other energy.Because of the complexity of the coal mine environment,most of the underground coal mines use belt conveyor for coal transportation.However,in the process of coal transportation,foreign bodies mixed in coal such as gangue and iron not only cause the scratch and tear of the belt,but also affect the quality of the coal.Therefore,this thesis takes how to effectively identify the foreign bodies in the process of belt transportation as the research object,and uses the related methods of deep learning to study.The purpose of this thesis is to discover the foreign bodies in the process of coal transportation in time,so as to effectively avoid the damage of foreign bodies to the belt and reduce the economic loss caused by foreign bodies.The research in this thesis is as follows.How to deal with the image data effectively is studied,in order to ensure the accuracy of subsequent recognition.Due to the characteristics of uneven light distribution,large dust and low contrast in the coal mine,the quality of the pictures is poor,which affects the subsequent recognition results.Therefore,image preprocessing is needed.Through the analysis and comparison of Histogram Equalization and Adaptive Histogram Equalization,Adaptive Histogram Equalization is selected for image enhancement.The method of image preprocessing is determined as follows: firstly,Median Filter is used to reduce noise,then Adaptive Histogram Equalization is used to enhance image.After preprocessing,the image is labeled and the data set of foreign bodies recognition is constructed.How to effectively identify the foreign bodies in the process of belt transportation is studied.According to the comparison of various recognition methods of deep learning,it is found that the image recognition based on Faster R-CNN algorithm is superior to other methods in speed and accuracy.Therefore,this thesis uses Faster R-CNN algorithm for image recognition,and constructs and improves its structure.Firstly,when building CNN,VGG16 is selected as the feature extraction network by comparison.Secondly,aiming at how to select the best region proposal for RPN generation,this paper introduces the idea of penalty function of confidence degree based on the traditional non-maximum suppression.Finally,the loss function is determined.Through the analysis of the processing results,it is found that the improved Faster R-CNN maintains high recall rate and accuracy rate.In terms ofrecall rate,gangue and iron respectively reach 97.73% and 95.00%,At the same time,the accuracy rate of the two is respectively 81.13% and 81.25%.And it has a fast detection speed to ensure the real-time detection process.Combined with the actual case analysis of C coal mine,calculating the economic benefits of using foreign bodies identification method based on Faster R-CNN to C coal mine.The economic benefits are calculated from three aspects: belt maintenance,reducing staff numbers,and Green mining.Through quantitative analysis,it is found that the method based on Faster R-CNN to identify foreign bodies in belt transportation is adopted in the coal mine,which brings better economic benefits to the coal mine every year.At the same time,the corresponding suggestions are given from three aspects: improving the relevant equipment,introduction of professionals,and improving organizational management.To ensure that the foreign bodies recognition based on deep learning is effectively used in coal mines.There are has 30 figures,9 tables and 90 references in this thesis.