Study On Obstacle Detection And Obstacle Avoidance Of Mobile Robot In Complex Indoor Environment

In the indoor environment, the existing obstacle detection system and the existing obstacle avoidance system of mobile robot has some defects such as obstacle detection effect is not ideal, easy to collide, complicated in calculation and so on. For the security requirements of mobile robot in indoor environment, to optimize Kinect blind spot detection, using a single Kinect sensor, and the method of panoramic obstacle detection depth image and machine vision to improve artificial potential field method, reduce the probability of collision and lock, the applicability of the mobile robot test platform was built to test our methods of obstacle detection and obstacle avoidance. The main researched work are as follows:1. In order to optimize the method of Kinect blind detection, the depth image rotating from the original pixel 640×480 to 480×640, which reduces the Kinect sensor to detect the effects on detection blind area of ground obstacles and suspended obstacle. The obstacle detection area can make sure that every part of robot body:Calculating the depth value after image flip extraction of formula to get the actual distance of obstacle.2. In order to obstacle contour of depth image is smooth and easy to detect obstacles pixel contour, the image is preprocessed; The background subtraction based on statistical average eliminates the influence of background in depth image:Then the obstacles in the image after background subtraction is separated from depth image:pixel area calculation of the obstacle image, filtering out the pseudo obstacle caused by noise; then by detecting the obstacle boundary, the number and value of the extracted boundary obstacles, which prepares data about obstacles for the next obstacle avoidance path selection.3. Calculating the actual position of the obstacle and boundary information to identify the local coordinate of obstacles relative to robot, convert the local coordinate of obstacles into world coordinates, the coordinates of the obstacle are taken as the obstacle position coordinates the artificial potential field method; Filtering of mobile robot obstacle regions, will not constitute a threat to filter out of robot obstacle and obtain the effective information of obstacles in order to reduce computation; The use of machine vision scene detection are processed failure scenarios of artificial potential field, the local minimum value selection of scene detection and obstacle avoidance path is proposed.4. Build a test platform for testing the method in the paper. The results show that the single Kinect sensor platform can detect obstacle robot range, and can successfully avoid obstacles; the improved artificial potential field method can avoid the defects of local minimum, walking along the wall method also greatly reduce the walking pattern invalid path, solved in narrow channel shock the problem, effectively realize the indoor robot in a dynamic environment and obstacle avoidance.