Obstacle And Slope Terrain Detection Based On Multi-Sensor Fusion

The autonomous operation and other functions of unmanned platform is highly dependent on the environmental information obtained by the perception module.The accuracy of environmental perception is an important factor affecting the safe operation of the unmanned platform,and obstacle size and slope terrain detection are important components of environmental perception of the unmanned system.At present,the research on obstacle information detection is generally categorizing obstacles,and there is little research on obstacle size detection.Although there is research on obstacle information detection based on the three-dimensional point cloud clustering algorithm,the detection error is large.At present,for the research of slope angle detection,the binocular image matching method is adopted,which involves large amount of data calculation and poor real-time performance.Or the slope gradient is calculated by fitting the slope equation based on three-dimensional point cloud.The methods above detect the relative angle of the slope,and the position information of the platform itself is rarely considered.In view of the shortcomings of the current research,this paper uses camera,IMU(inertial measurement unit)and 3D lidar as environmental perception equipment,and mainly studies are as follows:Firstly,in order to solve the defect of large error of obstacle size detection based on3 D point cloud clustering method,an obstacle information detection algorithm based on3 D point cloud and image fusion is proposed.For point cloud data processing,the ground point cloud filtering algorithm based on ray is designed and the Euclidean point cloud clustering algorithm is improved.The results show that the clustering effect is significantly improved compared with the original clustering algorithm.On the basis of point cloud clustering,based on the joint calibration results,the clustered point cloud is projected onto the imaging plane,and the obstacle pixel difference information is obtained through image processing.The obstacle size is calculated based on the proportional relationship between a single pixel and the size.The results show that the error of the proposed fusion algorithm for obstacle size detection is no more than 3.4%.Secondly,in order to make up for the defects of RANSAC slope fitting algorithm in detecting slope terrain and improve the accuracy of slope detection,a slope terrain detection algorithm integrating three-dimensional point cloud,IMU and image data is proposed.Firstly,the IMU and camera are calibrated jointly,and the coordinate relationship between any sensor is obtained based on the calibration results between the camera and lidar.For the problem of slope detection,a slope detection algorithm based on the fusion of 3D point cloud and IMU is proposed.By analyzing the coordinate transformation relationship,the 3D point cloud data is transformed into IMU horizontal coordinate system,and the slope is calculated based on IMU data and 3D point cloud slope fitting.On the basis of obtaining the slope angle,the slope area is extracted based on Lab color space.The slope gradient and the slope area extracted from the image are integrated to judge whether the slope is passable.The results show that the error of the proposed method for slope detection is less than 3.53%.Finally,indoor and outdoor obstacles and slope terrain experiments were carried out based on the proposed algorithms.The results show that compared with the single sensor detection method,the detection effect of the proposed fusion algorithms is significantly improved,and the detection errors of obstacle size and slope are both less than 4%.