Based On The Single Road Features Of Laser Radar Detection

Road feature detection in the complex environments is one important research content of outdoor intelligent robot autonomous navigation. Because the amount of data is small and the scaning accuracy is limited for the single line lidar, it is difficult for the small and medium-sized intelligent robots which only loaded the single line lidar to perform well for the problem of road feature detection. Hence, multi-frame interfusion for single-line lidar data is performed to solve the problem. Based on the spatial and temporal correlation characteristics between the multi-frame lidar data, the environment information within the larger area can be obtained, and then the robot can detect the feature of the road environment by this information.This paper focuses on the road feature detection technique based on multi-frame single-line lidar data fusion. The research area of this paper including dead reckoning method based on circular motion, combined positioning method of GPS and DR, multi-frame interfusion for single-line lidar data, correlation analysis among multi-frame lidar data, linear fitting and curve fitting for road boundary, obstacle detection based on height raster maps, traversable area extraction.The main research results of this paper are as follows:In order to solve the problem that the accuracy of the traditional DR positioning method is not high enough, this paper designed a DR positioning method which based on circular motion instead of straight line, this method can improve DR positioning accuracy effectively.For the problem of road boundary detection based on singe-frame lidar data, a dual threshold segmentation method is designed. And this method process with self-adapting selection of the dual threshold. Constraint relationships are constructed based on correlation analysis among multi-frame lidar data, and the experimental results show that, by the constraint relationships, this method can enhance the detection accuracy of the boundary points effectively.The problem of road boundary fitting is processed by using the least squares method. For solving the problem that the linear least squares method cannot fit the vertical distribution of boundary points, the least squares method based on the axis conversion is designed. For solving the problem that the curve road boundary points are sparse, the method of weighted least squares method based on segmented is designed.For the problem of obstacle detection, this paper propose to use the histogram to extract the feature of the road-surface area and then divide the obstacle area by this feature. This method can solve the segmentation problem that the obstructions target is too small and the obstructions feature is not obvious. The chain-code tracing algorithm based on the obstacle area is used to obtain the contour information of the obstacle. In the last, in order to obtain more effective information of the obstacle, the detected contour information is improved.For the problem of traversable region extraction, an algorithm based on regional growth is designed. In this algorithm, the current robot location is viewed as a seed and the height difference of adjacent raster area is viewed as the regional growth rule. Experiments show that the algorithm can handle a variety of problems that traversable region extraction in the complex environment.