基于改进的Otsu法对红花图像分割的研究  

Study on safflower image segmentation based on an improved color difference model

作　　者：董芙楠 郭辉 潘江如 韩英杰 DONG Fu-nan;GUO Hui;PAN Jiang-ru;HAN Ying-jie(School of Control Engineering,Xinjiang Institute of Technology,Urumqi Xinjiang 830039,China;School of Mechanical and Electrical Engineering,Xinjiang Agricultural University,Urumqi Xinjiang 830052,China)

机构地区：[1]新疆工程学院控制工程学院,新疆乌鲁木齐830039 [2]新疆农业大学机电工程学院,新疆乌鲁木齐830052

出　　处：《林业机械与木工设备》2023年第8期68-74,共7页Forestry Machinery & Woodworking Equipment

基　　金：新疆维吾尔自治区重点研发计划项目“特色经济作物收获机械装备研究及集成示范”(2018B01001-1)。

摘　　要：红花图像分割的准确率直接影响后期红花空间定位的精度。利用改进的色差模型结合Otsu法,建立一种准确分割目标红花的方法。首先,将RGB颜色空间转换为HSV颜色空间,利用H、S、V阈值范围清除图像背景像素,初步对图像像素进行筛选;然后利用双边滤波法对筛选后的图像进行预处理,随后利用基于最小二乘原理的色差模型,快速准确分割目标红花;最后,采取最小外接矩的方法标记出目标红花和重心点,将标记效果融合到红花原图中,完成红花识别。结果表明,基于改进的色差模型图像分割效果明显,准确率高达90%,对于复杂背景的红花图像具有一定的适应性。The accuracy of image segmentation of safflower directly affects the accuracy of spatial localization of safflower.In order to establish an accurate segmentation method of target safflower,an improved color difference model combined with Otsu method was used to achieve real-time segmentation.Firstly,the RGB color space was converted into HSV color space,and the background pixels of the image were cleared using the threshold range of H,S and V,and the image pixels were initially screened.Then bilateral filtering method was used to preprocess the screened images,and then color difference model based on least square principle was used to quickly and accurately segment the target safflower.Finally,the minimum external moment method was adopted to mark the target safflower and the center of gravity point,and the marking effect was integrated into the original image to complete the safflower recognition.The results show that the improved color difference model is effective in image segmentation,with an accuracy of up to 90%,and has certain adaptability to safflower images with complex background.

关 键 词：红花 分割 OTSU法 色差模型 最小二乘原理 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]