Design And Implementation Of Blind Anti-collision System Algorithm Based On IMU And Vision

People with visual impairments in China are a large number,and they have many inconveniences in their daily lives,especially when they travel.With the continuous advancement of science and technology,there are also many products that help the visually impaired people to travel,such as guide shoes,guide glasses,etc.,the principles are various,but there are still many defeats in actual use,such as high cost,poor real-time performance and can't provide rich environmental information for the blind.According to the above situation,this thesis developed the anti-collision system for the blind based on the visual and inertial test unit(IMU),which realizes the three functions of obstacle recognition,stair recognition and voice warning,and can provide real-time road condition information for the visually impaired people to avoid danger by voice.The main research contents of this thesis are as follows:(1)By using the sparse point cloud data which is obtained under the current position of the camera based on the monocular visual inertial system,analyze the distribution of the sparse point cloud in the three-dimensional space,and construct a suitable obstacle recognition area with the camera as the center to identify the obstacle.The area is divided into three parts:the left side,the middle part and the right side Judge whether there is an obstacle in the location by calculating the number of point clouds in it which is specified in the obstacle recognition area,and remind the visually impaired person of obstacles that may appear in this location by means of voice warning.(2)Due to the special structure of the stairs,it is not only an obstacle that impedes walking but also an access between stairwells for visually impaired people.It cannot be simply seemed as an obstacle and needs special consideration.In this thesis,based on the LSD linear detection algorithm and the characteristics of the stair steps,the straight line of the stair edge in the complex environment can be screened out,and the existence of the stair can be judged according to the identified linear distribution and the number of straight lines.Finally,according to the recognition result,the system will remind visually impaired people by the voice warning.(3)By analyzing the daily travel roads,designing representative experimental scenes according to the distribution of obstacles that may be encountered,comprehensively testing each function in the anti-collision system,and adjusting the optimization system according to the specific results,and designing the interface to improve the user-friendliness of the system.The function of the whole system is implemented in Linux system.The OpenCV computer vision library and the Ceres nonlinear optimization library are used as the development environment.Finally,the scene of the simulated road condition is used as the experimental environment to test and optimize the anti-collision system.The test results show that the system fits in with the design expectations and can meet the daily travel needs of the visually impaired group.In the process of simulating blind walking,it can be found out that the obstacles and stairs are accurately identified,the warning is timely,and the extensibility is strong.