嵌入式全景视觉实时目标识别

Embeded Omni-vision Object Recognizing in Real-time

在嵌入式系统中,使用鱼眼镜头实现全景视觉,完成实时目标跟踪工作。采用FPGA+DSP+PC的硬件系统架构作为识别器,使用SAA7113H芯片采集图像,再由FPGA将采集后的图像存入到SRAM,因为使用鱼眼镜头采集后的图像是畸变的,所以采用FPGA对图像进行矫正,运用Freeman链码进行目标识别,尤其是对直线的识别,这些工作都需要大量的计算,高效的FPGA和Altera公司提供的IP核,加快目标识别速度,以达到目标跟踪的目的。

This paper presents an embedded system,which depends on fisheye lens with omni-vision real-time in order to object recognizing.The FPGA（field programmable gate array）＋DSP（digital signal processor）＋PC（private computer） provides a hardware platform for on-board recognizer,which involves collecting image utilizing SAA7113H chip and then stores the processed image into SRAM（static random access memory）.Since we employ fisheye lens,the image will be distorted,making use of FPGA to correcting the distorted image,and exploding the rreeman chain code methods to recognize the object,especially aiming at line recognizing.These tasks need amounts of calcutation and employ the high-perform of FPGA and IP（intellectual property） core from Altera,which accelerate the speed of calculation and reduces the time of object recognizing,it is better for object tracking.