Trees Used In Robot Navigation And Positioning Detection Algorithm Research

The mobile robot localization and navigation(SLAM) is the premise of the robot’s realization of autonomous operation and other specific functions, so it makes great sense and is challenging. The purpose of SLAM is to build environmental map by the environment perception systems such as sensors and relevant data processing methods, in which it is a crucial task to choose and detect landmarks. When the robot is in the outdoor environments, even in the forest environment, we often choose trees as landmarks. In this paper we study the tree detection algorithms in the light of robots which is applied to the unstructured environment contains trees, in order to lay a foundation for the landmark detection in visual SLAM.. The main research contents are as follows:1) we study about an new environment recognition algorithm in the light of advantages and disadvantages of bottom-up and top-down visual environment recognition algorithm. we mainly use a Steerable Pyramid Filter to gain the feature vector, and then using it to train the geometric classifier, at last using the classifier to classify the environment, laying the foundation for the landmarks choose and detection.2) we propose tree appearance-based tree detection algorithms. Firstly we proposed the tree detection algorithm basis on the fusion of texture and color, in this algorithm we fuse the color and texture to gain the new feature vector, we train the classifier with the new feature vector and classify tested images with the classifier to recognize the tree trunks in the setting; Secondly we use the brightness instead of color to improve the operating speed and fuse the brightness and texture for object detection; Thirdly we try a statistical region merging algorithm. Test the three algorithms and verify their performances.3) we propose a structure-based tree detection algorithms. we detect the tree contour lines, work on the screening on them to delete the wrong lines and realize the detection and the tracking of the trees. we test the algorithms using Matlab and verify its performance. The comparison of the structure-based algorithm to the appearance-based algorithm is carried out to reflect the superiority of the former algorithm.