基于背景分离与亮度校正的苹果表面缺陷提取.docxVIP

基于背景分离与亮度校正的苹果表面缺陷提取 Abstract： Apple has the highest yield of fruits in our country， but the export volume is less than % of the total output. Fruit surface defect detection technology based on image processing can promote the classification and sale of fruits. This article takes Red Fuji Apple as the research object. Firstly， the RGB image of Apple is obtained， and the HSV model of the image is obtained based on the conversion formula. The threshold segmentation and hole filling of the saturation S component are used to obtain the apple background separation template. Then the background separation map of the R component is obtained according to the background separation template， and the R component after the separation of the background is subjected to luminance correction. Finally， the single threshold segmentation method is used to extract the defects. A total of apple samples were collected during the experiment， of which were defective and were normal. The defect recognition rate of the defective fruit reached .%， and the discrimination rate of the normal fruit was %. The experimental results show that the S-component acquisition background separation template combined with the brightness correction algorithm can obtain better defect extraction results without adding additional cameras. Key Words： defect extraction; background separation; brightness correction; threshold segmentation 引言 截至年，我國水果总产量已达到 万吨，然而总出口量只有万吨，不足总产量的%，经济效益不够明显。与其它水果出口量占比较大的国家相比，主要差距表现在果品的采后处理环节。果品采后处理是果品增值的关键，采后处理一般包括催熟、分级、打蜡及储存保鲜等步骤[]。其中水果分级技术是果品采后处理的核心技术，也是目前亟待解决的技术难题[]。水果分级的依据主要是颜色、尺寸、形状以及有无缺陷，与其它指标相比，表面缺陷检测一直是果品分级中的重点及难点[]。苹果是一种四季畅销的大众水果，含有矿物质、糖酸、维生素等多种营养物质[]，深受广大消费者喜爱，是我国十分具有代表性且年产量最高的水果[]，也是我国出口量最大的水果之一，因此对于苹果表面缺陷检测技术的研究意义重大。 传统图像处理方法和深度学习方法在水果表面缺陷检测中都有应用。如Heinemann等[]利用RGB图像信息计算出H分量，并依此识别出苹果表面的果锈;Leemans等[-]基于苹果的彩色图像信息，分别采用高斯分类器及贝叶斯分类方法对苹果图像的每个像素进行分割，从而识别出缺陷部分;朱伟华等[]通过RGB彩色模型获得HSL彩色模型，并基于此提出模糊颜色聚类方法用于缺陷分割;李锦卫等[]将马铃薯的RGB图像转换为亮度指标，并利用灰度截留分割法快速找出可能的缺陷区域