Obstacle Detection Method Based on Single Camera
|School||Harbin Institute of Technology|
|Course||Control Science and Engineering|
|Keywords||Mobile Robots Obstacle detection MRF Image Segmentation SCM|
From the 1960s on a mobile machine SHAKEY first birth , and now a new generation of Boston invention can walk on snow BigDog, mobile robots are widely used, has become a hot spot of the world's research . A mobile robot obstacle avoidance focused research lies . Only by achieving obstacle avoidance, people are able to move the machine in an uncertain environment secure mobile . This article describes a novel based on monocular vision for obstacle avoidance obstacle detection methods. When the mobile robot detects an obstacle , it is possible to effectively avoid it and is safe to move. First, we will apply machine learning methods to obtain high斯马尔科夫random (Gaussian MRF) model, which mainly reflects picture links between features and depth . When the model is known , after , as long as the picture input to the model can be obtained depth map of the picture . Secondly , the joint segmentation of images and predicted depth map obtained obstructions in the picture position. Calibration matrix is then passed into a position in the world coordinate system obstacle deflection direction . When the mobile robot to know the distance and direction of the obstacle , then it can effectively avoid obstacles . In this paper , laser scanners used to capture the depth of training information, data acquisition system of the mechanical structure and hardware are designed for faster and more efficient data collection. Finally, the simulation results show that this algorithm can effectively detect the depth and direction of an obstacle for the mobile robot obstacle avoidance great help .