Measurement Of Coal Quantity In Coal Mine Skip

In the demand of automation,machine vision has a wide range of applications in various fields.The working environment of underground coal mine is very complicated and harsh.It is dangerous for workers to work in the mine.Long-term underground work is not good for health,so it is very difficult to recruit workers in the mine.In order to reduce the workload of workers,realize the automatic operation of coal mine and reduce the artificial pressure of coal mine,this thesis applies structured light and deep learning to the actual work of coal mine.Structured light 3D reconstruction and convolution neural network are widely used in many fields at present.This thesis has completed two tasks: the center extraction of coal light stripe in coal mine skip based on structured light and the monitoring of skip working state based on convolutional neural network.This thesis first introduces the technical method of extracting the center of skip coal strip.The actual work is divided into skip work image acquisition link,image preprocessing link and light strip center extraction link.After the skip work video is collected in the field,the image is preprocessed to reduce the workload of subsequent image processing.In the entire system,the most crucial point is to accurately extract the center of the linear structured light stripe projected on the surface of the object.Currently,there are many methods for extracting the center of a laser beam.In this thesis,several classical center extraction algorithms are applied to the work of a coal mine skip.Experimental results show that Steger algorithm has great advantages over other algorithms,so Steger method is selected as the final extraction algorithm.An improvement is made based on the Steger algorithm which can greatly increase the computational speed of the system by combining the Steger algorithm with a method based on RGB color threshold segmentation.And at last,the useless information will be processed by introduced them to a connected domain.The finally light strip extraction will be relatively accurate which can meet the experimental objectives.Secondly,this thesis introduces the technical method of skip working state detection.The method based on deep learning is used to identify the working status of skip.Res Net50 neural network is used as the training model,nearly 7500 images were selected as the data set and the final accuracy rate is 0.987%,which is relatively high accuracy rate achieved to identify the working status of skip.Finally,the trained model is applied to the actual working video of the skip,and it can meet the requirements of actual working conditions.In this thesis,based on structured light,the research on the coal sliver center extraction of coal mine skip and the monitoring technology of skip working state based on convolutional neural network has achieved good results in the experiment.It can accurately extract the sliver center of skip coal and accurately classify the working state of skip,and time the unloading time of skip.