基于多尺度多方向分解的自然场景下麦穗计数  

Wheat ears counting in natural scenes based on multi-scale and multi-direction decomposition

作　　者：鲍文霞[1,2] 张婷婷[1,2] 胡根生 梁栋[1] BAO Wenxia;ZHANG Tingting;HU Gensheng;LIANG Dong(National Engineering Research Center for Agro-Ecological Big Data Analysis&Application,Anhui University,Hefei 230601,China;School of Electronics and Information Engineering,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学农业生态大数据分析与应用技术国家地方联合工程研究中心,安徽合肥230601 [2]安徽大学电子信息工程学院,安徽合肥230601

出　　处：《安徽大学学报（自然科学版）》2020年第6期20-27,共8页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61672032,41771463);农业生态大数据分析与应用技术国家地方联合工程研究中心开放课题项目(AE2018009)。

摘　　要：小麦是我国重要的粮食作物,准确的小麦麦穗计数是小麦产量精确估计的前提.针对自然场景下小麦图像的频域分布特点,提出了一种多尺度多方向分解的小麦麦穗计数方法.首先,该方法根据待处理的小麦麦穗图像中麦穗和背景信息分布的频段不同的特点,对小麦麦穗图像进行多尺度多方向分解,获取能突出小麦麦穗信息的多尺度多方向子带,减小土壤、小麦叶片等背景信息的干扰;然后,利用灰度阈值分割方法对小麦麦穗子带图像进行分割,利用形态学中的膨胀和腐蚀等操作实现包含小麦麦穗信息的连通区域的分离;最后,利用Find maxima计数方法实现小麦麦穗计数.实验结果表明,该方法对小麦麦穗计数的精准度明显优于其他基于颜色特征和纹理特征小麦麦穗计数方法.Wheat is an important food crop in China,and accurate counting of wheat ears is the prerequisite for accurate estimation of wheat yield.Aiming at the frequency domain distribution characteristics of wheat images in natural scenes,a new method of wheat ears counting based on multi-scale and multi-direction decomposition was proposed in this paper.First,according to the different frequency bands of the wheat ears and background information distribution in the image of wheat ear to be processed,the image was decomposed in multi scale and multi direction to obtain multi-scale and multi-direction sub-bands that could highlight wheat ear information,thereby reducing the interference of background information such as small soil and wheat leaves.Then,the gray threshold segmentation method was used to segment the wheat ear sub-band image,and the morphological operations such as expansion and erosion were used to separate the connected areas containing wheat ear information.Finally,Find maxima counting method was used to realize wheat ears counting.Experiments were developed and the results demonstrated that the accuracy of our proposed method for wheat ears counting was significantly better than the other wheat ears counting methods based on color and texture features.

关 键 词：麦穗计数 多尺度多方向分解 特征提取 形态学处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]