基于Bayes的竹片颜色检测分级方法

为了采用机器视觉对竹片自动识别与颜色分选,研究了一种基于竹片图像颜色特征与纹路特征和Bayes分类器的颜色分类方法.首先,对灰度图像采用Canny算子进行边缘检测,再利用Hough变换对竹片进行边缘定位,并对倾斜竹片实施旋转校正,以确定待检测竹片在图像中的具体位置.根据竹片的位置提取竹片区域平均颜色特征及纹路特征,将其作为样本的属性特征,采用Bayes训练的颜色等级作为输出,建立特征参数与颜色等级之间的Bayes分类器,上位机获得分级信号后经串口通过下位机实现竹片的自动分级.试验结果表明,该方法对竹片颜色检测准确率达到91.7%,可为竹制品行业的竹片颜色自动在线检测提供理论依据.

Bayes-Based Color Detection and Classification Method of Bamboo Slices

Using the color and texture characters of bamboo slices and Bayes classifier,a new color classification method was studied to realize sliced bamboo's automatic recognition and color classification based on machine vision.Through Canny operator the edges of the sliced bamboo in gray image were detected,and then using Hough transformation the edges were located.For those edge slopes that were not vertical in the image,rotation revise was performed to get their exact locations in the image,based on which the characters of mean color feature and vein image feature were extracted as the character features of samples.With Bayes-trained color classification grades as the output vector.the Bayes classifier was set up between mean feature values and color grades.After the host computer obtaining the classification signal,the auto classifying of bamboo slices was realized through slave computer subsystem by the serial port.The results show that classifier's precision rate is 92% for color detection of bamboo slices.The system realizes the on-line automatic recognition of bamboo slices' color.