摘要
针对油菜籽粒图像阴影会直接影响后续籽粒分割和计数的准确性问题,基于油菜籽粒外观纹理特征提出了一种基于图像中心旋转的多边形扫描阴影去除算法。以油菜黔油早125作为研究对象。采用剖面线分析其籽粒纹理特征,获取目标和阴影灰度分布规律及分布范围。而后对多边形区域纹理特征分析,结合灰度分布范围约束,最终实现阴影消除。对140张油菜籽粒样本图像进行测试,结果表明:该算法处理每张图像平均耗时1.90 ms,能够很好地去除单粒、连接籽粒之间的阴影,较基于行的灰度统计阴影去噪算法去除阴影更加彻底、准确,鲁棒性更强,虽然处理速度低于前者,仍满足实时性要求。
Concerning the problem that the image shadow of rapeseed grain directly affects the accuracy of subsequent grain segmentation and counting,a method that shadow denoising was proposed based on the appearance texture characteristics of rapeseed grain.The rapeseed Qian You Zao No.125 was used as the research object.The grain texture characteristics were analyzed by hatching to obtain the grayscale distribution law and distribution range of target and shadow.Then,the texture feature analysis of the polygon region was combined with the grayscale distribution range constraint to finally achieve shadow elimination.Total of 140 images of rapeseed samples were tested.The experimental results show that an average of 1.90 ms was consumed for processing each image,and it is a good way to remove single shadows and connected shadows between grains.Compared with the line-based grayscale statistical shadow denoising algorithm,the proposed shadow removal algorithm is more thorough,accurate and robust,and although the processing speed is lower than the former,the real-time requirement is still satisfied.
作者
彭顺正
岳延滨
李莉婕
赵泽英
孙长青
PENG Shunzheng;YUE Yanbin;LI Lijie;ZHAO Zeying;SUN Changqing(Guizhou Agricultural Science and Technology Information Institute,Guizhou Academy of Agricultural Sciences,Guiyang Guizhou 550006,China)
出处
《计算机应用》
CSCD
北大核心
2019年第S02期247-252,共6页
journal of Computer Applications
基金
贵州省星火计划项目([黔科合成转字(2015)5336号])
贵州省农业科学院青年基金资助项目([2018]56号)
关键词
油菜籽
纹理特征
阴影识别
阴影定位
阴影去除
rapeseed grain
texture feature
shadow recognition
shadow positioning
shadow removal
