A Obstacle Avoidancing And Navigation System For Visually Imparied Based On Computer Vision

The statistics shows that there are about two hundred and eighty-five million of people suffered from visual impairment,of which thirty-nine million are blind and two hundred and fortysix million have low vision in different degrees.As one important part of human sense,vision plays a very significant role in people's daily life.People will be incapacitation for basic movement and recognition if they lose their visual sight,so visual impairment become the concerns of the society.The existing assistant product for visual impairment mostly are single-function,low convenience and slow update.Along with the continuous developments and breakthroughs,most of the technologies in computer vision obtained optimized,including camera calibration,image recognition and matching,three-dimensional reconstruction,vision enhancement and so on.So it becomes a hot trend to apply computer vision to visual aids for visual impairment people.Simultaneously,along with the development of hardware,it becomes realizable to embed complex image processing algorithm with transportable and handy mobile device,which can reduce the burden of people with visual impairment.This paper come up with a new obstacle avoiding and guidance system for individuals with impairment by combining the computer vision technology and emerging TPVM(Temporal Psychovisual Modulation)technology.The system mainly solves the common problem in the life of the visually impaired,such as that it is hard for them to avoid the obstacles and recognize the status of the pavement.In this paper,we propose some algorithm solution to solve this problem.And finally we probe into the algorithms and experiment with the theory to confirm that the solution is effective and available.The main algorithms proposed in this paper includes:Firstly,monocular distance measurement algorithm base on feature matching,the algorithm combined the monocular vision technology and feature point matching to measure the distance of the obstacles.By comparing the position relation of feature points in different images,the depth of the feature points can be determined,and according to the depth information,we can get the distance of the obstacles in the image.The mainly feature matching algorithm includes SIFT,SURF and so on.We research the principles of these algorithm and experiment with them.This distance measurement algorithm is used to avoid obstacles for people with visual impairment.Secondly,the dual-view projection system using temporal psychovisual modulation,which can hide information.TPVM is an ingenious interplay of signal processing,optoelectronics,and psychophysics which exploits the psychovisual redundancy in refresh rate of the modern optoelectronic displays.So high refresh rate display device can provide much more information than the amount that human visual system can obtain.Based on this theory,it is achievable that the synchronized camera and human eyes can get different information from the save image scene projected,so some of the information is hidden.Thirdly,the recognition algorithm for the status of pavement based on structured light.We used the TPVM projection system to hide the structured light to avoid the discomfort feeling to people' eyes.And then by processing the image set captured by camera,we will give a feedback to the users about the status of the pavement.The algorithm specifically includes feature extracting,ID3 algorithm and so on.And finally,the system will feedback the status of the pavement by voice.The system proposed build a entire frame which is aimed for the visually impaired base on the algorithm mentioned above,the system can aid the visually impaired to avoid the obstacles and recognition the status of pavement,and finally some of the function is transferred to the mobile device,which promoting the convenience of the system.