Application Analysis Of Depth Extraction From Single Monocular Image

In the researches of object recognition, path planning and scene interpretation, depth information has an important use. Humans can easily estimate how far away from the obstacles when we walk, but for clumsy robots, it becomes too difficult to judge how far away objects are. Along with the robotic navigation technology has become the focus of the mobile robot industry, how to estimate the distance has become an important task. Depth estimation has variant methods by using camera and image processing, which in robotics and computer vision mostly uses stereo vision. Stereo vision is fundamentally limited by the baseline distance between the two cameras, and it may suffer from sudden changes of image intensity due to the illumination noise, also using two cameras makes the cost much expensively. Thus we consider the task of depth estimation from a single monocular image.We first introduced some basic theories of machine vision and Gaussian-Markov theorem which we will use to build model. On the basis of human s vision system, we build the probabilistic model. To collect images and their depthmaps, we designed the laser scanner system. At last, we trained the parameters of the model and test for new images. The result proved that this method can estimate depth information from monocular images, and can be used for robot navigation.