Research On Blind Obstacle Avoidance Technology Based On Binocular Vision

According to the World Health Organization survey data,there are currently about 285 million visually impaired people in the world,and 39 million are completely blind.There are about 12.63 million visually impaired people in China,and there are great challenges in travel.The traditional methods of guiding the blind,such as guide dogs and guide sticks,are costly and unsafe,and less research has been done on intelligent electronic guide devices to avoid obstacles.To address this situation,this paper focuses on the research of blind travel obstacle avoidance technology.The details of the research are as follows:(1)The investigation of general scheme of blind obstacle avoidance technology is studied.This paper proposes a multi-sensor combination of intelligent obstacle avoidance scheme in combination with current obstacle avoidance technologies.The blind person’s near-range obstacle avoidance technology based on binocular vision is determined,using binocular vision to obtain scene information to generate an obstacle avoidance map,and to carry out walking path planning,using distance measurement sensors to measure the distance in real time to ensure the safety of the blind person’s travel,so as to achieve the effect of reasonable obstacle avoidance and safe obstacle avoidance;(2)Research on 3D information acquisition and map reconstruction technology of binocular vision for proximity obstacle avoidance for blind people.The binocular vision system is used to obtain the parallax information of the scene and get the 3D coordinates of the points in the scene.The post-processing algorithm of parallax map combined with weighted least squares filtering and grayscale hole filling is used to make the effective information rate more than 80%,and the improved otsu algorithm(chunking secondary threshold method)is used to segment the obstacles and fuse them with the parallax map.obstacles within 15 meters ahead,with relative error of position less than 8%;(3)Research on blind obstacle avoidance path planning correction algorithm.The algorithm introduces the cost of crossing to solve the problem that the traditional A* algorithm cannot plan whether to cross or not,uses the turning penalty factor to suppress turning,and can reduce the turning rate by about 30% in the same scene.The safety of obstacle avoidance is increased,and the above algorithms are integrated to output obstacle avoidance commands,so that the blind can reasonably avoid obstacles;(4)Safety and security module research.The ultrasonic distance measurement and infrared distance measurement sensors are selected to detect obstacles in the 300 cm and 80 cm areas in front and side of the blind person respectively,and the detection area is divided into three levels,and the corresponding instructions are outputted to prompt the blind person according to the different danger levels,so as to solve the problem that the blind person has large safety hazards in the process of obstacle avoidance due to incomprehensive scene shooting and travel deviation;(5)Information scheduling and overall testing.Obstacle avoidance and emergency command priority settings are made according to the degree of importance of the information,and the generated information is output prompted by using the speech synthesis module for hierarchical classification.The hardware experimental platform is built using Raspberry Pi 3B+for joint software and hardware debugging,and the obstacle avoidance function is tested in real scenarios.The results show that the binocular vision-based blind obstacle avoidance technology studied in this paper has good applicability and can reconstruct the scene in front of the blind and plan the path correctly,and establish a safety zone 300 cm in front of the blind and 80 cm to the side to further ensure the safety of the blind and meet the needs of the blind for obstacle avoidance.