Research On Blast Hole Identification Of Intelligent Explosive Filling Robot And Passable Area Planning

At present,in the process of underground mine blasting in China,the filling of explosives is completed by manual or simple injection equipment.In this kind of working environment,not only the air is humid and the environment is bad,but also the working efficiency of workers is low,the labor intensity is high and extremely dangerous.Therefore,it is not only the urgent need to improve the intelligent level of mine blasting equipment,but also the key link to upgrade the intelligent industry in the mining industry to use robots instead of artificial operation in the explosive filling process.This topic mainly deals with the problem of visual recognition in the subject of “robot automatic filling explosives”.The main research contents are as follows:First,the development of underground emulsion explosive blasting device is introduced,and the development and research status of emulsion explosive mixed loading vehicle device at home and abroad are introduced and explained.In view of the traditional explosive filling device,the intelligent design is carried out,and the overall design scheme is determined again.The scheme includes grab design,bracket design,vehicle control,recognition algorithm,hand eye system and artificial intelligence platform.This paper mainly focuses on the realization of recognition algorithm of borehole.Secondly,this paper mainly analyzes the method of robot automatic hole searching and injection based on depth learning,in which the depth camera collects the position of the blast hole in the mine and combines it with the robot end motion trajectory planning.The identification of blast holes is decomposed.The traditional image recognition method is used to design the target recognition algorithm.By extracting color,texture and shape features,and then classifying the extracted features,the results show that the traditional image recognition algorithm has poor adaptability.For example,the recognition accuracy of the no uniform features such as brightness,size and angle is low,furthermore,this paper studies how to use the depth neural network to identify the blast hole.Thirdly,this paper fully compares the network structure of RCNN and Faster R-CNN.By comparing the accuracy and speed of recognition results,Faster R-CNN algorithm used in this paper is determined,and the algorithm is introduced and studied in detail.The RPN network,the core part of the algorithm,is decomposed and discussed.In order to fully prove that Faster R-CNN algorithm is more suitable for the working environment of identifying blast holes in underground mines,two otherrecognition algorithms are compared,YOLO and SSD512.Under the requirement standard of borehole identification,this algorithm can process 5 images in one second,and the accuracy is higher than that of YOLO and SSD512,and the accuracy can reach over 97%.Moreover,Faster R-CNN algorithm will not fail to identify small objects.Finally,under the established underground mine artificial environment,the algorithm of blast hole recognition and robot end passable area and obstacle area planning is verified.The results show that the recognition algorithm and control method used in this paper can effectively realize the work of explosive filling.