为研制仿生眼球,构建了一个嵌入式的基于仿生控制的视觉图像处理系统。根据仿生眼对视觉系统的要求,设计了一个以TMS320DM642 DSP为核心,以TVP5150、SAA7121H为编解码模块的视觉图像处理系统;实现了仿生眼视觉识别的一系列软件开发,包括实时图像采集、视频图像处理、视频输出及目标位置参数传递等功能;在CCS2.2的环境下,对视觉识别算法进行了探索,用经典的Prewitt算法以及改进的Prewitt算法对系统进行实验测试。
In order to develop a biomimetic eye, an embedded video image processing system based on biomimetic control is established and the hardware, software design process as well as the algorithm design is investigated. Firstly, in accordance with the requirements of the biomimetic eye visual system, a system based on TMS320DM642 DSP, and used TVP5150, SAA7121H for the codec modules is designed. Then after developed a PCB successfully, a series of software development based on visual recognition is realized, including real-time image collecting, video image processing, video output, parameters transfer and so on. Finally, in the environment of CCS2.2, visual recognition algorithm is explored, using classical Prewitt algorithm and improved Prewitt algorithm for the test.