A Video Vehicle Contour Detection Algorithm Based On Improved Adaboost Algorithm

As the key technology of the Intelligent transportation,the video vehicle detection technology based on contour has a broad application prospect.Also,the Intelligent transportation has the stringent requirements to the video vehicle detection technology based on contour,such as real-time and the accuracy.The main question is the all kinds interferences in the background of the video when the detecting.With the domestic and foreign scholars' continuousefforts,there is a lot of methods about video vehicle detection technology based on contour came out.The Adaboost algorithm is the popular Machine learning algorithm in this years,and it has a good performance on Face detection field.Adaboost algorithm also can apply to field of video vehicle detection technology.This paper put forward the improvement based on the classic Adaboost algorithm.The reaseach work as follows:1,introduce the background meaning of the video vehicle detection technology based on contour,know the questions in the video vehicle detection technology.2,Introduction and analysis the existed algorithms about video vehicle detection technology,understand the theory and sums up the problems.3 introuduce the Adaboost algorithm in detail,understand the thory and how to use it.introduce the Haarfeatures,Integral figure and how to training and chose the classifier.4,put forward a algorithm which improve the Adaboost algorithm on the video vehicle detection technology based on contour.First,put forward to tailor the training sample,because the haar features has a large amount of computation,it needs a lot of time.Cut of the edge pixels can reduce the number of the featuresand lower the computation.5,when the Adaboost algorithm is running on the video pictures,the sliding window will detect the picture from left to right,up to down,the irrelevant information of the picture was also detected,it costs time.In this paper,use the optical flow method to get the moving regions as the region of interest.Detect the edge of the ROI with canny algorithm,filter the ROI through the edge energy.In the end,use the classfier to detect.6,Set a threshold value to filter the result.Looking forward the future combine the advantage and the weakness of the algorothm.