Research On The Unmanned Environment Sensing System Of Mining Electric Locomotive

The coal mining industry is closely related to national energy security.my country has begun to vigorously promote the intelligent construction of coal mines.The era of unmanned and intelligent coal mine auxiliary transportation has arrived.The unmanned coal mine auxiliary transportation electric locomotive can transport materials and personnel more safely and efficiently.The electric locomotive environmental perception system senses and analyzes the surrounding environment in real time to identify the position and status of various elements such as roads,obstacles,and pedestrians.As an important part of unmanned electric locomotive driving,a stable and reliable environment perception system is an important guarantee for the realization of unmanned electric locomotive technology.In light of the above situation,this thesis designs an environmental perception system for mining electric locomotives,with the following main research contents:(1)Investigate methods for jointly calibrating and fusing data from Li DAR and camera systems,with the goal of improving the accuracy and consistency of the collected data.The ultimate aim is to provide more precise three-dimensional environmental perception information.Firstly,the joint calibration of sensor space is achieved by using camera parameters and corner point pairs,and data is synchronized using timestamped frames.Then,the 3D point cloud data is projected onto 2D space,and the correspondence between 2D point cloud data and camera images is established by comparing their coordinates in 2D space.(2)The detection method for obstacles is investigated.Preprocessing of data obtained from Li DAR includes distortion removal and noise reduction of point clouds.The Li DAR point clouds are then subjected to ground segmentation using the polar grid algorithm.Then the point cloud on the ground is used to extract the rail information using the method of image and point cloud information fusion,and the driving area is determined by filtering the sidewall point clouds on both sides of the roadway in the straight as well as the curved roadway of the well through the rail boundary.The optimized dynamic threshold Euclidean clustering algorithm with density spacing is used to cluster the obstacle point clouds and box out with the minimum enclosing box,and then obtain the geometric features and location coordinates of the obstacles.(3)Research on object tracking algorithm based on obstacle detection.Introduce the working principle of JPDAF(Joint Probabilistic Data Association Filter)algorithm to track the obstacles by combining data association and state estimation.To address the shortcomings of the original algorithm,the algorithm efficiency is improved by filtering out poorly correlated events and reconstructing the confirmation matrix;the filter is made to have better adaptive performance by adding the forgetting factor.The tracking manager scheme is designed to fit the scenario,and the target state is maintained and updated in real time.(4)Experimental validation of the proposed method is conducted using real-world environmental datasets.The experimental dataset is based on data collected from the electric locomotive experimental platform in a school coal mine environment.During the experiment,a point cloud clustering and segmentation algorithm is used to verify obstacle detection and tracking using the researched algorithm.The experimental results are compared and analyzed to verify their stability in practical applications.