The Recognition System For The Blind Way Based On Image Processing

The building of blind ways can be seen as an inevitable project in municipal engineering. Nowadays, blind ways have been lay on pavements in almost all of the cities. However, it becomes a common phenomenon that blind ways can't give their full plays. This thesis points out that the reason may be the various blind ways have not been built according to the standards or man-made destroyed which impaired their guiding function heavily. The purpose of this thesis lies in structuring a set of complete blind way recognition system based on image processing, which can collect the images of the location of the blind persons when they are on blind way on the real-time. Furthermore, this thesis employs various algorithms about image processing to recognize the location and direction of the blind ways and makes comparison with the walking direction of blind persons, then makes interactional modifications on the real-time.Due to the great differences between blind ways, the single method of image segmentation can hardly recognize all the blind ways correctly. So, this thesis suggests a unified processing framework which uses the methods of image segmentation - image edge detection - blind way edge searching in turn in order to detect the edge of blind way and confirm the relative location of blind persons.Hence, this thesis contains the following parts:1. The structure of the whole system is introduced which includes hardware system, software system. The core part of this system, namely, the methods of unified recognizing various blind ways based on image processing, is focused.2. The present situation of the blind ways surveyed is introduced. The images of blind ways are categorized and relative image segmentation methods are made. They can be categorized into the three main types: the method based on threshold segmentation, the method based on texture differences segmentation, the method based on the internal geometric features of the blind ways. Among them, the method based on texture differences is a difficult point in recognition. So this method is emphasized which includes various extraction methods of feature vectors and clustering method. Besides, a clustering method based on graph is brought out.3. In order to extract blind way edge, this thesis employs Canny edge detection operator to analyze the results of image segmentation at first. Then, it employs Hough transform to search the straight lines according with the condition. Furthermore, Freeman chain code and an algorithm based on it are introduced. The algorithm which can detect the location of corner points in edge detection results is pointed out.In the conclusion of this thesis, the experiments results of the images of various blind ways are given which manifests that this method can recognize the regions and directions of blind ways in a prompt and effective way. Basically, it can be concluded that this method has reached the standard of practical application...