基于显著性轮廓的苹果目标识别方法

发布时间：2018-03-09 04:16

本文选题：图像分割　切入点：苹果　出处：《计算机应用研究》2017年08期 　论文类型：期刊论文

【摘要】：正确地将苹果从图像中识别出来是苹果采摘机器人实现自动采摘的前提,在对该问题研究的基础上,提出一种基于显著性轮廓的苹果目标识别方法。利用K-means无监督聚类算法将图像分割为背景和目标区域;由于光照等因素,目标区域内部存在大面积空洞,引入ASIFT特征,将完整的目标与存在空洞的目标进行特征匹配记录与空洞相对应的特征,由这些特征恢复成像素填补空洞;在基于区域的基础上,采用g Pb轮廓检测器进行轮廓检测,生成较长、较明显的灰度轮廓图像;通过动态阈值Otsu法对灰度轮廓图像进行自动阈值处理,去除目标周围大量的边缘噪声,确定连续的显著性轮廓,完整地提取目标轮廓。实验结果表明,该方法具有更好的准确性与鲁棒性,对苹果目标的正确识别率在98%以上。
[Abstract]:Correctly recognizing apple from the image is the premise of automatic picking of apple picking robot. Based on the research of this problem, A method of apple target recognition based on salient contour is proposed. K-means unsupervised clustering algorithm is used to segment the image into background and target area. Due to illumination and other factors, there are large areas of holes in the target area, so the ASIFT feature is introduced. The complete target is matched with the target with holes and the corresponding features of the hole are recorded. These features are restored into pixels to fill the void. Based on the region, the gPb contour detector is used to detect the contour. A longer and more obvious gray contour image is generated, and the dynamic threshold Otsu method is used to process the gray contour image automatically to remove a large amount of edge noise around the target and to determine the continuous significant contour. The experimental results show that this method has better accuracy and robustness, and the correct recognition rate of apple target is more than 98%.
【作者单位】：常州大学信息科学与工程学院;常州轻工职业技术学院;
【基金】：国家自然科学基金资助项目(51307010) 江苏省自然科学青年基金资助项目(BK20140265,BK20140266) 常州市科技计划资助项目CJ20160010)
【分类号】：TP391.41

【相似文献】