Research On Obstacle Detection And Self-Exploration Of Indoor Robot Based On Computer Vision

With the continuous advancement of technology,mobile robots are playing an increasingly important role in our lives.They are helping humans to complete important tasks such as escort,excavation,transportation,rescue,exploration,and investigation.Mobile robots are gradually moving from semi-autonomous operations to fully automated operations.The mobile robot's autonomous exploration technology has become the basis for robot movement in an unknown environment.Robot autonomous exploration means that in the environment without prior knowledge,the robot needs to sense the surrounding local environment information through the robot's own distance sensor,model and analyze the environmental information,and make continuous motion position decision.Achieve the purpose of exploring the entire environment.All the research in this paper is aimed at independent exploration.The main research contents include:Firstly,the related problems of indoor robot exploration are briefly introduced,and the environment modeling method is briefly introduced.The common robot exploration methods at home and abroad are listed.In the self-exploration,the robot needs to identify the location of the feasible area and the obstacle.The robot detects the position of the obstacle in the environment through its own depth camera.The depth camera can obtain the color map and depth map of the environment.Image segmentation and edge detection of the segmented image.Obtain the edge of the object in the foreground and background in the environment.For the depth map,it needs to expand and remove the noise,then obtain the approximate position of the obstacle in the environment through threshold segmentation,and finally obtain the obstacle in the environment by correcting the processed color map and depth map.The specific location of the object and the test in the actual environment.The environment in which the robot is located is modeled based on information theory.The long-term exploration goal of the robot is to minimize the information entropy of the map.The short-term goal is to maximize the information gain.The position with high information gain is equivalent to the unexplored area of the environment.The first Bayesian estimation is used to estimate the information gain distribution of the entire map,and the extreme position of the information gain in the map is calculated.In the process of continuously moving to the position where the information gain is the largest,the robot is driven to explore the unknown area.The software and hardware platform of indoor mobile robot was built,and the experimental method based on Bayesian optimization was carried out using the built mobile robot platform.The number of exploration steps,time and exploration of traditional methods and Bayesian estimation were compared.The difference in integrity,the experimental results show that the Bayesian optimization-based exploration method can make the robot continuously explore the unknown region without a priori information,and compared with the traditional method,Bayesian optimization method The estimation of environmental modeling is more accurate,the exploration time is shorter,the number of steps is less,and the integrity of the exploration of the environment is at an intermediate level.