中文摘要：

以TI公司的多媒体图像处理器TMS320DM643为核心处理器，提出了一种大米外观检测视频处理嵌入式系统硬件的实现方案．基于MATLAB平台对大米样品的视频系统回读图像分别进行了Sobel算子及其改进算法的仿真比较研究．结果表明，改进算法可以检测出更为完整的大米边缘，实现了大米视频图像的实时检测，处理一帧图像仅需42．7ns．

英文摘要：

We proposed an embedded system hardware implementation scheme for rice appearance detection video processing by using a multimedia image processor TMS320DM643 of TI company as the kernel processor. Firstly, the read-back images of a rice sample video system were respectively simulated by using Sobel operator and the improved algorithm thereof based on the MATLAB platform. The simulation results showed that the improved algorithm could detect more intact rice edge. Then, the improved Sobel edge detection algorithm was selected in the hardware system to achieve the real-time detection of the rice video images, wherein one frame of the image could be processed in only 42.7 ns.