Design And Realization Of The Image Tracking Based On Embedded System

Image processing with embedded system is a challenge to the traditional image processing hardware. Combining image acquisition, image processing and image display in one system, its implementation will open a new way for image processing, and new field for the application of embedded systems. Meanwhile, due to a wide range of image processing applications, the system itself has a wide application prospect.This dissertation presents the design and implementation of image tracking embedded software system with an embedded hardware platform based on TE2410. Considering the basic functions of image processing system and the characteristics of embedded system design method, summering the characteristics of ARM-based embedded system and the latest developments, with the advantages and disadvantages of multiple operating systems in mind, we choose embedded Linux operating system as the platform. We ported it to the ARM920T processor-based hardware platform, constructed the embedded Linux cross development environment. with the characteristics Linux driver architecture in consideration, we introduced the principles and procedures of how to collect, display and store the image data in embedded Linux circumstances, complete the image preprocessing module, image processing modules, and graphical display interfaces. Preprocessing module mainly realizes image acquisition program using USB-based camera, display program, and storage procedures. In the image processing module, this paper involves pretreatment which is suitable for real-time requirements of this task. In image segmentation, we use Otsu threshold method in a fixed time to obtain a dynamic threshold in response to environmental change. In the time interval, a fixed threshold segmentation method is used to improve efficiency; then through Blob analysis the segmented image, the location of target image, target area, perimeter, centric and a series of geometrical features are given. Comparing these geometric features with geometric characteristics of the target we realize target recognition. Each frame of the image sequence in the target chain is tagged, with the purpose of tracking the moving target.In order to achieve optimal performance, this paper using Open CV algorithm for image processing module simulation to improve the part which has an impact on performance to meet the demand of real-time processing.