基于BP神经网络的图像识别研究  

Image Recognition Based on BP Neural Network

作　　者：万来毅[1] 陈建勋[1] 王卫平[1] 李俊[1] 

机构地区：[1]武汉科技大学计算机科学与技术学院,湖北武汉430081

出　　处：《武汉科技大学学报》2006年第3期277-279,292,共4页Journal of Wuhan University of Science and Technology

摘　　要：提出一种采用BP神经网络实现钢材编号文字识别的方法。先采用彩色图像HSI空间中S分量的特性,定位钢材区域,然后利用一系列图像处理技术,对图像中钢材编号区域定位、分割字符,最后采用BP神经网络进行字符识别。实践证明,采用BP神经网络,可有效地识别钢材编号,速度快、识别率高,具有较高的实用价值。A method for the recognition of serial numbers on rolled steel based on BP neural network is discussed in detail. First, the rolled steel area is located based on the color picture＇ s S weight＇ s trait of HSI space. Then, the serial number regions from rolled steel area are divided and characters segmented after a se- ries of image processing techniques is used to the picture. Finally, BP neural network is applied for character recognition. It＇ s demonstrated by practice that BP neural network can be used for fast and efficient recognition of serial numbers on rolled steel with high accuracy and practical value.

关 键 词：BP神经网络 区域定位 HSI空间 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]