实时图像识别系统  

A Real-time Image Recognition System

作　　者：沈定刚[1] 戚飞虎[1] 李艳斌[1] 

机构地区：[1]上海交通大学光纤研究所

出　　处：《高技术通讯》1996年第7期37-40,共4页Chinese High Technology Letters

基　　金：国家自然科学基金

摘　　要：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。A real- time image recognition system is reported in this paper. Object recognition algorithm is composed of Knowledge Based Image Segmentation, Wavelet Transform Based Rotation Invariant Feature Extraction and Neural Network Based Classifier. Hardware architecture contains dominant computer,intel80860 Chip and image I/O & preprocessor. Technical property of the hardware architecture is better than Mark III and close to Mark IV, both of which are produced by TRW company. This system can recognize objects from the complicated background.

关 键 词：图像识别 神经网络 图象通信 

分 类 号：TN919.8[电子电信—通信与信息系统]