The Research Of The Multi-Objective Image Segmentation Algorithm In Complex Environment

Computer intelligent video surveillance system is used for coach bus can realize its operational status of the effective supervision. Currently, although some coach bus are installed surveillance cameras, but the actual monitoring task still need a lot of artificially work to complete, and now the most of the information which video monitoring system provided is without any analysis data, which can't give full play to its intelligence, real-time supervisory role. The research of the multi-objective image segmentation algorithm in complex environment is use to segmentation the multi-objective image in video image sequences from the video monitoring system. It's the base of follow-up processes, such as: the target's feature extraction and identification of the target, tracking and behavior understanding. It's the key technology of the intelligent video surveillance system to realize the play their intelligence, real-time supervisory. It is also a efficient method to governance the driver who get the passenger charges in midway and not reported. This thesis main research work and creative achievements show below:(1) Aiming at the Hough transform's problem which required large storage space and the low speed of the calculate, we improve it. First we change the radius uncertain round detection into the radius determines round detection. Secondly adopts maximum method and the round does not cover principle, standard deviation method to avoid threshold judgment, Improve the detection accuracy and speed.(2) Using an method which combining an improved 2d OTSU method and quantum particle swarm optimization to segment the multi-objective image in complex environment, and improved the segmentation accuracy and speed.(3) Puts forward a kind of new methods for support vector machine sample selection. Under the guidance of the 2-d gray-level histogram to choice a certain number of samples to train SVM, using it to multi-object image segmentation in complex environment. Not only got the ideal segmentation effect, but also improved the segmentation speed.(4) Extracting the texture characteristics, wavelet characteristics and geometric characteristics from the segmented image, and using euclidean distance method to achieve the overhead identification. Further verify the efficiency of the algorithm...