Study On Bus Stop Recognition Algorithm Based On Machine Vision

With the rapid development of machine vision and multimedia technology, the machine vision and multimedia technology has been widely applied to solve practical problems in production and life. With the increase of city population and the number of private cars, bring a series of serious problems to the urban traffic. The development of intelligent transportation, improve the urban traffic intelligent and information technology is a new solution to solve the traffic problems. To improve the public transport service quality, promote the development of bus, is an important way to solve urban traffic problems. The machine vision and video information processing technology is applied to the bus station system, not only can realize the information and intelligent of the bus station system, but also in accordance with the theme of "intelligent transportation", has great practical significance.This thesis mainly use the bus automatic station system as a background, study bus station identification algorithm based on machine vision. Through the camera capture the video along the bus route in real-time, use key frame extraction technology extract multiple key frames before the bus station, before this, establish image database which save the pictures before the bus station, match the key frames and images in the database. After a successful match then drive the bus-stop system to report the station.The whole system use machine vision technology and image processing technology for real-time processing of the collected video. When extract the key frame, use the time information eliminate large redundant fragment. Then make a crude key frame extraction along the bus routes based on the content key frame extraction. When one frame matching success, use key frame difference clustering extraction method to extract the key frame each 10 s, extract the key frame which is needed. After key frame extraction, key frame image and image in database need feature extraction and matching. Considering the special situation of bus station, this thesis proposes the image contour Harris-SIFT algorithm. First extract the contour of the key frame and image in the database, and uses the Harris corner select feature points of the image, which combines Harris algorithm and SIFT algorithm, get the feature points of the image contour, also improving the feature point descriptor, calculate the Euclidean distance between the descriptor then match images. Use the guided complementary matching method and voting filtration method to improve the matching accuracy.In this thesis, according to the characteristics of automatic bus-stop system, according to the characteristics of human vision is also proposed to extract the bus routes on a region of interest, according to the interest part of text area and traffic signs in the bus routes were extracted and recognition, extract the interest region can also assist the bus station judgment, improve intelligence and accuracy of the bus-station.