基于机器视觉的禽蛋图像分割方法研究  

Study on egg image segmentation method based on machine vision

作　　者：贾超[1] 郁志宏[1] 王亮[1] 武智慧[1] 温鹿[1] 

机构地区：[1]内蒙古农业大学机电工程学院,呼和浩特010018

出　　处：《东北农业大学学报》2011年第5期77-82,共6页Journal of Northeast Agricultural University

基　　金：内蒙古自治区自然科学基金项目(2009MS0906);内蒙古自治区高等学校科学研究项目(NJ09050)

摘　　要：探讨了一种新的图像分割算法,可对禽蛋表面图像中脏斑、血斑等信息进行高效、准确的分割。文中将由CCD摄像机获取的禽蛋图像首先采用改进的中值滤波方法去除噪声,然后将去噪后的图像使用Sobel边缘检测算子检测边缘,运用分裂合并算法对去噪图像进行多区域分割,从而得到较为明显的多区域图像,最后利用图像融合技术将Sobel算子检测到的边缘图像和区域分裂合并算法得到的多区域图像融合为一体。与其他方法相比,采用上述方法可以获得边缘较为清晰,脏斑、血斑等表面信息更为明显的分割图像。经试验验证该种方法实时性较好,分割结果令人满意,可以应用于禽蛋品质的在线检测。This article presented a method that could efficiently and accurately detect and segment egg image surface information such as dirty spots,stains and so on.The image taken by CCD camera was denoised with improved median filter.And then the denoised image was detected with sobel operator and apparently segmented multi-regional images through splitting and merging algorithm.Finally,the two new images were syncretized to form a new image by image fusion techniques.The algorithm above could obtain segmented images with legible edge and clear surface information of dirty spots and stains.The results showed that real-time of the segmentation algorithm was good and the segmented results were satisfactory,which provided necessary pretreatment to the egg quality detection.

关 键 词：分裂合并算法 边缘检测 图像融合技术 

分 类 号：TP391.14[自动化与计算机技术—计算机应用技术]