Development Of Mining Truck Loading Rate Detection System Based On Deep Learning And Image Recognition

Carrying measurement is a daily production management work of the mine.There are the following shortcomings in the current transportation measurement and detection methods:(1)Manual statistics.The method of multiplying the number of transport vehicles by the transport volume of each vehicle is used to make a rough statistics of the mining volume of ore and rock.This not only causes serious distortion of open-pit mine production data and affects the accuracy of ore blending,but also causes waste of labor,vehicles,and fuel.(2)Weighbridge.It is easily affected by bad weather such as rain,the equipment is easy to be damaged,the failure rate is high,and the installation in the underground is not conducive to maintenance and after-sales maintenance.(3)Sensor detection.Sensor equipment has a high cost and a high failure rate.At the same time,it is greatly affected by severe weather such as rain and snow,as well as dust and finely divided slag in the mine,which brings inconvenience to daily maintenance.(4)Level meter detection.The level gauge detection method requires multiple level gauges and control instruments to be used together,which requires a high response time,and at the same time,it is more difficult to measure the volume of ore in a sheltered transportation tool.Based on the above background analysis of the shortcomings of the open-pit mine transportation measurement method,finding a low-cost,fast and intelligent method for detecting the transportation measurement method of mining trucks is of great significance to the rational mining and utilization of ore.This paper designs and develops the subject of a mining truck loading rate detection system based on deep learning and image recognition.The aim is to apply artificial intelligence and image recognition to the detection of the loading rate of mining trucks,and to improve the stability and automation of the existing mining truck loading rate detection.It is helpful to optimize the production operation and scheduling plan of the mining enterprise,and finally realize the goal of improving the efficiency of open-pit mining,improving the utilization rate of equipment and improving the intelligence of the mine.The main work of this paper is as follows:(1)Image acquisition and preprocessing.The production process of the self-built data set is described,and the simulated mine cart images in the laboratory environment and the real mine cart images in the open-pit mine environment are collected and made into a TFRecord format data set.In order to further optimize the detection effect,three image preprocessing strategies of wavelet threshold denoising,bilateral filtering,and histogram equalization are selected to prepare for the subsequent model training process.(2)Construction of volume detection model.In order to realize the effective detection of mining truck loading any volume,this paper proposes to combine the VGG16 classification network model in deep learning with the least square regression model,and build the mining truck loading rate detection model which combines VGG16 and the least square method.The model realizes the detection of mining trucks with arbitrary loading rate,and has good generalization performance.(3)Model training and verification.This paper uses a large number of simulated minecar pictures in a laboratory environment and some pictures taken at the open-pit mine as the training set of the model.Using 10 types of laboratory environment simulation mine car pictures and 4 types of open-pit mine field shooting pictures as the test set and supplementary test set of the model,to verify the detection accuracy and generalization performance of the model.Research experiments have proved that the mining truck loading rate detection model has a good detection effect.The detection errors of the mining truck model in the laboratory environment and the real mining truck in the open-pit environment are 5.95%and 4.21%,respectively.(4)System development.Starting from the analysis of the system function,the paper develops a load rate detection sstem for mining trucks by training the VGG16 deep neural network to train the optimal network parameters and the best fitting parameter input model in the least square method.The development of the system greatly saves manpower and material resources,improves the efficiency of loading rate detection and statistics of mining trucks.The whole detection system has good applicability and real-time operation,and can realize the visualization of the detection results of the loading rate of mining trucks,which has a good guiding significance,research significance and application value.