The Research Of Forestry Obstacles Identification Based On 3D Laser

In order to meet the requirement of automation and intelligence of forestry machinery and reduce the labor intensity, this paper studied the automatic identification about forestry obstacles. After the point cloud data was denoised and simplified, chose appropriate method to segment it. Then extracted the characteristics of the four kinds of targets. Based on the extracted characteristics to achieve the targets identification. This study provided theoretical basis for the research of forestry automation and intelligence.The main research work is listed as the following:(1) Point cloud data preprocessing. Removed the outlier of point cloud data based on Statistical Outlier Removal filter of point cloud library(PCL). Then simplified the data by using Geomagic Studio software. The data preprocessing provided research basis for the point cloud segmentation and recognition.(2) Point cloud segmentation. A new point cloud segmentation algorithm based on feature fusion was proposed according to clustering point cloud segmentation algorithm. Segmentation was realized after fusing the advantages of normal vector and laser reflection intensity. And compared this segmentation results with that from algorithms based on Euclidean distance clustering.(3) Chose and extracted characteristics of tree, stone, floor and person. The characteristics included height probability distribution, point feature and laser reflection probability distribution.(4) Based on the feature data, did the point cloud recognition by using support vector machine algorithm and BP neural network algorithm respectively, and optimized the algorithm.