A Guide Blind System Design Based On Digital Image Processing

The sidewalk of big cities are all have blind lanes,but many blind lanes have destroyed or occupied seriously,and it is difficult to have actual effect.Current information guide system for blind people based on GPS and sensors to identify,which is more depended on server workstation,the equipment is complex,high cost and may be lost contact in the area away from server.Therefore,this paper presents a guide system mainly based on embedded system and image recognition,the guide device is made into a form of portable blind crutch which could reduce the cost and does not rely on the server that can achieve self guide for blind people.In this paper,The author gives the overall scheme of hardware design of the guide system and hardware circuit.The first is image acquisition by a CCD camera installed on the cane,and then is image pre-processing filtering on FPGA platform.Development work on the FPGA platform mainly in the following three parts: Firstly,is to achieve SAA7113 control and data interface on the FPGA,followed by the data communication with the PXA270 processor platform through the FPGA SDRAM read and write,and the third is in the filtering of the digital image pre-processing on FPGA platform.The filtered image is identified by the image processing algorithms in the PXA270 processor for guiding,and the results transform to voice prompts to guide blind by a speech output module,if the image processing algorithm anomalies and can not guide the blind successfully,it can be avoidance seeing by ultrasonic distance sensor attached to the cane.In this paper the author expounds analysis and design process of image processing algorithms about the guide system.The first is blind image feature detection,after comparing SIFT feature detection and SURF feature detection,we used a faster SURF feature detection as the detection algorithm,which lead the blind into the middle of the blind lane.And then using Canny operator to extracted edge image,after that is to use the Hough transform to process the edge image and recognize the straight path,and helping ensure the blind walk in the middle of the blind lane.While multi-feature comparison judgment joint decision algorithm to determine whether there have any obstacle or not.Using the captured the image as a background image by the camera in the blind cane,once in a while,a current image be captured as a test image.The background image and the image treatment measured by comparing extracted every variance,correlation and edge density as three characteristics.After training to determine the threshold,jointdetermination can determine whether the obstructions ahead.If there have an obstacle,they are prompted seeing around the obstacle,after re-testing to find and import to blind lane by SURF feature algorithm.At the end of the article,the author tested the every module of the guide system,and gives the test results.Tests show that the system in most cases can effectively determine the blind road and can successfully guide,with a success rate of more than 80% of the guide.