Vehicle Detection And Tracking Based On DM6437

The era of rapid development in the computer science and mathematics, machine visionhas been widely used in many fields, including the people’s daily life, industrial production,and even national defense and other important areas. An important application of machinevision technology is the intelligent monitoring, accurate detection is one of the primaryproblem for fast moving object. After years of research, one of the moving object detectionhas made great progress, and provide theoretical support for the to further intelligent controlsystem. The same time, DSP and other high-speed processor provides hardware support forembedded system implementation of complex algorithms.Through the use of intelligent transportation systems, we solve the increasingly serioustraffic problems, the detection and tracking of vehicles compared to provide technical supportto the realization of this goal. PC-based vehicle detection and tracking algorithm has beenincreasingly exposing the lack of not only costly, but also in image processing speed, powerconsumption is also far lagged behind the DSP. DSP dedicated signal processing hardwarestructure and pipeline technology so that it has a great advantage in terms of signal processing.With the development of the times, people are required to have smaller, more cost-effective,smaller power systems, which require us to proceed at a more advanced vehicle detection andtracking system, the DSP platform, compared with a good choice. Gaussian backgroundmodel based TMS320DM6437hardware platform, corrosion expansion, connectedcomponent labeling algorithm, to achieve real-time vehicle detection and tracking system, themain hope in the follow-up work of the Intelligent Transportation Systems traffic statistics,accident analysis, analysis and identification of the license plate and lay a solid foundation.This article describes the commonly used moving object detection method based theTMS320DM6437build hardware development platform, the design software developmentplatform, focusing on research TMS320DM6437single Gaussian background model basedalgorithm to achieve. The main contents are as follows:Based on the background of the single Gaussian model motion object detection methodis the basic detection method, this paper background model of established by a statisticalmethod determining the respective pixels of the Gaussian distribution; separation of thebackground point and before Attraction of using the threshold value discriminating France;background model updated EM method to update. And in accordance with the real-timecharacteristics of the embedded devices, the method of the initial background model, usingthe own modify views of the first frame as a Gaussian model, the mean value of the currentpixel value, the variance of the initial assignment, the background model update and the separation of the background and the foreground are carried out simultaneously, immediatelywhen a point pixels of the current frame is judged to be the background profile parameters, asa model of the detection of the next frame, when is judged as a foreground and backgroundseparated as prospects, and its corresponding background parameter is not changed byTMS320DM6437platform, indoor and outdoor environment, the actual running, showed thatthis system can be provided in the video at least10~15/S is used, real-time effective vehicledetected and tracked vehicles, with a more high accuracy and robustness.