论文部分内容阅读
基于图像目标识别的智能性及实时性要求，设计出了一种实时图像识别系统。图像目标识别算法由知识型图像分割、基于子波变换的旋转不变性参数提取和神经网络分类器三部分组成；硬件系统由主控计算机、Ｉｎｔｅｌ８０８６０芯片、图像输入输出与预处理模块三部分组成，硬件系统的性能指标超过ＴＲＷ公司的ＭａｒｋⅢ，接近ＭａｒｋⅣ。采用此系统实现了对非均匀光照或复杂背景下目标的实时识别。
Based on the intelligence and real-time requirements of image object recognition, a real-time image recognition system is designed. Image target recognition algorithm consists of three parts: knowledge-based image segmentation, rotation invariant parameter extraction based on wavelet transform and neural network classifier. The hardware system consists of the main control computer, Intel80860 chip, image input and output and preprocessing module, The performance index of the hardware system surpasses Mark III of TRW Company, approaching Mark Ⅳ. The system is used to realize the real-time recognition of targets under non-uniform lighting or complex background.