Study On Algorithm Of Object Detection In Intelligence Video Surveillance System

With the projects of "Security China" deepening gradually, the network video monitoring system has become increasingly prominent role in the maintenance of social security and fighting against the crime. With a lot of real-time/non-real-time videos, the traditional monitoring methods which are dependent on manual observation and distinguish can not suffice the requirements, we need some algorithm system which can analyze the object's behavior and feature in video frequency automatically, and the moving object segmentation and tracking technology have applied maturely in the intelligent monitoring. Because the object detection technique is relatively complex in the requirement of efficiency and engineering application, it is not used in present network video monitoring system almost. So this study takes the object detection technique as the important study content, and provides the surveillance system a sort of object detection algorithm which adapts the general image quality and scene structure, and integrates the moving object segmentation and object tracking technique to provide algorithm groundwork for target behavior analysis and fine feature extraction.The fundamental principle and algorithm of HOG (Histograms of Oriented Gradient) and SVM (Support Vector Machine) technique are analyzed by this study. The detection algorithm based on HOG has been proved which has enough robustness and good results of detection. HOG detection method need vast and representational image sample to train classifier, but the same object's HOG is different in different visual angle. So it is the urgent key problems that transform the sample's HOG from the certain visual angle to another visual angle in application of HOG detection method in practical projects. This paper proposes a transition algorithm for the sample's HOG in different pitching angle, rotation angle of object and rotation angle of optical. It can be to improve the results of SVM classify and the accuracy of HOG algorithm, and reduce the quantity of positive and negative sample for training.The GMM algorithm and Mean-Shift object tracking algorithm are also analyzed by this study, GMM and Mean Shift algorithm are taken as the moving object segmentation and tracking technology, therewith the study constructs an intelligent monitoring algorithm system which includes scene information constructs, moving object segmentation, object detection and tracking. In order to realize this algorithm system, this study uses the modulation design method, and uses the opened source class OpenCv and software of VC++6.0 to develop a experiment platform, and realizes the video image input, moving object segmentation, object detection, object tracking and data output by the platform.At last, 1180 of the images are selected as positive training examples with nearly same viewpoints, together with their left-right reflections (2360 images in all), and 9580 of the images are selected as negative training examples, and we do detection on more than two hundreds of video database (different static environments) which include 7694 human with the platform. Moreover, this study also does vehicle detection with the platform. The tests show that the method is accurate and effective, and reduces the quantity of training sample for object detector training, and achieve anticipated effect.