Design Of Traffic Flow Detection System Based On Video

With the rapid development of Intelligent Transportation Systems, video detection technology is becoming more and more important in the transport industry. Most traditional video based detection methods make use of general-purpose processor, at the personal computer platform. There is much inconvenience when applied to traffic at the scene. With the improvement of information technology, embedded systems have been applied to many fields, which could make good use of the dedicated processor. Video detection system based on the embedded system has also become a hot spot in Intelligent Transportation.In order to achieve the purpose of vehicle detection, this paper makes use of TI's Davinci technology, detects the vehicles on embedded operating system. Dual-core processor deals with the system control and image processing algorithms respectively, which represents the traffic detection technology development. It is an innovation that using Davinci technology in Intelligent Transportation Systems field.This paper does some research on the traffic flow detection algorithms based on video after introducing the composition and processing of video detection system. In comparing several commonly used algorithms, the background subtraction algorithm that focuses on how to update the background is selected. This paper uses an adaptive algorithm for automatic updates of the background. After image preprocessing, binarization segmentation, background update and threshold update, moving objects can be detected clearly.In hardware design, according to the detection requirement, this paper uses a dedicated video processing chip TMS320DM6446 as the main chip, the ARM is responsible for system control, and DSP for algorithm treatment. In addition to basic video input and output modules on the video detection board, there are also many others, which include memory modules, real-time clock module, power supply and reset circuit and a wealth of communication modules, such as network interface, USB interface, RS-232, RS-485 and so on.At software of board support package, this paper migrates the popular open-source software u-boot as the main bootloader. In the three-stage booting mechanism (ROM bootloader + user bootloader + u-boot), the embedded operating system starts up. In order to fit traffic detection requirement, this paper does some clipping to operating system, and re-optimizes the kernel configuration, then constructs an embedded file system. Both the kernel image and file system image are downloaded to FLASH, the system could implement self-booting when powered on.Apply this detection algorithm to the embedded platform designed in this paper at different junctions, it works well and achieves satisfactory results in different weather conditions.