The Three-dimensional Environment Identification System Of Intelligent Excavating Machinery

With the rapid development of computer control technology and artificial intelligence technology,people’s demands for intelligent excavating machinery are also improving.In the process of actual excavation,excavator drivers often face the complex excavating working environment,working continuously under high intensity,not only bring huge load to the body,but also bring some security risks to the entire excavating process.The intelligent operation of excavating machinery enables people to change from manual labor to mechanical work,thus freeing people from heavy manual labor and improving work efficiency greatly.The 3D environment identification technology of excavating machinery plays an important role in excavating intelligent mechanical autonomous operation,and it is the basis of automatic control and autonomous movement of excavating machinery.Therefore,an intelligent,efficient and real-time 3D environment identification system for excavating machinery is designed in this paper.Through the real-time scanning processing of excavating machinery operation scene,the reliable identification of excavating target,delivery target and obstacles is realized.In the course of the thesis,the following work is done:The hardware platform of 3D environment identification system for intelligent excavating machinery is set up,including the selection of environment sensing equipment,and the selection of the turntable and the motor,and the connection of the hardware circuit.The software of 3D environment identification system for intelligent excavating machinery is compiled.The programming environment and developing tools are introduced,and the GPU parallel computing method based on CUDA is designed.A method for region segmentation of 3D cloud data for excavating machinery’s working scenes is proposed.The segmentation of 3D cloud data is performed by the idea of layer by layer segmentation.Firstly,we calculate the geometric features of 3D point cloud data,and extract the principal component features.Secondly,the excavating scene’s data is divided into three categories: point point,linear point and surface point according to the size of the feature.Then,constructing the ball of the principal component characteristics,the Meanshift clustering algorithm is adopted across the normal vector clustering and the DBSCAN clustering algorithm based on spatial location is adopted for further segmentation,and the two methods were optimized.Then,the final ground extraction method based on the normal vector direction of the spatial region segmentation and the position elevation is designed.Finally,the region segmentation of the 3D cloud data is carried out.An object classification method of 3D cloud data for excavating machinery’s working scenes is proposed.Firstly,the mathematical model of the standard excavating target and the delivery target is established,then the segmented spatial region fragments are matched with the standard model to achieve robust identification.Secondly,the data of the remaining excavation scenes are classified and identified by constructing a random forest model classifier,and the obstacles are also accurately identified from the actual scene.Finally,in order to facilitate the debugging and analysis of the staff,a human-computer interaction graphical interface based on OpenGL is designed.