Port Environment Based CENTRIST Feature Pedestrian Detection And Research

Image recognition technology is an important field of artificial intelligence, withdevelopment of various disciplines such as the computer vision, machine learning,pattern recognition, artificial intelligence, which has playing a more and moreimportant role in all aspects of national life. Importance of aritificial intelligent imagerecognition technology is becoming more and more prominent，which has apply to theentry of airports, customs entry for intelligent monitoring, automatic alarm for forestfiring, product testing of industrial production. Pedestrian recognition is an importantsubject of intelligent image recognition. It can effectively enhance the efficiency ofthe intelligent monitoring of sensitive area, improve the pedestrian filtrating of securitydepartment, at the same time a good detection method for understanding thesubsequent behavior of pedestrian. Yet state of the art pedestrian detection methodscan not have the combination of high efficiency and high accuracy.This paper is based on the practical intelligent monitoring project in TianJin portby Shanghai science and technology co. LTD. First expounding a pedestrian detectionsystem in the general, introducing the architecture of system, the acquisition oftraining data and the method of moving object detection. Second, focus on pedestriandetection based on characteristics of CENTRIST. The results show that we have got agood method of pedestrian detection. The main content:1) Intelligent video acquisition card access to background services program. withthe contact of constructing video acquisition card and the background system,realizes the multi-channel input. basically got the demand of the backendsystem2) Design and implement the pedestrian detection module with the practicalapplication value. According to the result of motion detection to extract the pedestrian operator, we have got a reliable results of classification with theneural network classifier and SVM classifier.3) Test the moving target detection algorithm and classifier of patternrecognition.with the comparative analysis of these algorithms, we got a moresuitable CENTRIST motion detection and classifier algorithm operator.Method described in this paper can also be used in other object detection such asdifferent types of vechie.