基于计算机视觉的竹块颜色分类方法研究  

Color Classification of Bamboo Pieces by Computer Vision

作　　者：张昭[1] 何东健[1] 

机构地区：[1]西北农林科技大学机械与电子工程学院,陕西杨凌712100

出　　处：《安徽农业科学》2010年第26期14751-14753,14783,共4页Journal of Anhui Agricultural Sciences

摘　　要：为了实现计算机视觉对竹块的自动识别与颜色分类,研究了竹块图像的颜色特征、颜色分类方法。首先,将原始图像由RGB空间转换为HSI空间,用OTSU法确定阈值对原图灰度图像分割背景,用radon变换倾斜校正,提取竹块颜色判定有效区域,并计算其HSI的均值和标准偏差,作为特征参数输入BP网络进行训练。结果表明,该方法简捷有效,人工神经网络与人工分级的平均一致度为94.5%。In order to recognize and classify bamboo pieces automatically,a detection system based on color classification through computer vision was proposed.Firstly,RGB color was converted to HSI color.Then,bamboo object was segmented with threshold estimated by Otsu algorithm.The tilt registration was carried out based on radon transform.Then the effective region for color classification was extracted and the mean and standard deviation of the pixels in the region for H,S and I component respectively were calculated as the color feature parameters,which were sent to BP neural network for training.The experiment revealed that the method was efficient,the average consistence between computer vision classification and manual classification reached 94.5%.

关 键 词：竹块 颜色 分类 神经网络 计算机视觉 

分 类 号：S56[农业科学—作物学]