【摘要】 针对蔬果分级系统,提出了一种基于主色直方图特征匹配的蔬果颜色分类算法.首先在OHTA颜色空间进行图像分割,确定蔬果范围,并通过改进的矢量中值滤波法滤除彩色图像中的噪音;其次将彩色蔬果图像由RGB颜色空间转化为符合人类视觉特点的HSV颜色空间,将蔬果颜色在HSV颜色空间非等间隔量化,通过变换将三维颜色矢量降为一维颜色矢量,产生新的索引颜色,通过聚类的方法提取蔬果的主色特征,并以颜色直方图表示。通过把被检测蔬果的主色直方图与各等级的标准蔬果图像直方图比较,实现对蔬果的分级.该分类方法实现简单,不必进行特定训练,简化了分类器的设计,具有一定通用性.试验结果表明,分级精度达到97%,满足了实用要求.更多还原

【Abstract】 The dominant color histogram matching method for fruits and vegetables classification was proposed.Firstly,through the image segmentation in OHTA color space, the fruit’s shape is separated from the image,and the noises in the color images are filtered through the improved vector median filtering algorithm.Then the color image is transferred from RGB color space to HSV color space which matches human’s vision,and the fruit’s colors are non-interval quantified to get new index colors.The dominant color features are picked up via clustering method,and represented in color histogram.By comparing fruit’s color histograms with the standard grading color histograms,the fruits are classified easily.In this way,there is no request for specific training on machine,and the design of classification machine is simplified.The experimental results show that the classification accurate rate is up to 97%,which can satisfy practical requirements.更多还原