苹果采摘机器人目标果实快速跟踪识别方法

为了减少苹果采摘机器人采摘过程处理时间,对苹果采摘机器人目标果实的快速跟踪识别方法进行了研究。对基于R-G颜色特征的OTSU动态阈值分割方法进行首帧采集图像分割,采用图像中心原则确定要采摘的目标果实;利用所采集图像之间的信息关联性,在不断缩小图像处理区域的同时,采用经过加速优化改进的去均值归一化积相关模板匹配算法来跟踪识别后帧图像的目标果实,并进行不同阈值分割方法实现效果,不同灰度、亮度和对比度的匹配识别以及新旧方法识别时间对比试验,从而验证了所采用和设计方法的有效性;其中所设计跟踪识别方法的识别时间相比于原方法,减少36%。

Fast Tracing Recognition Method of Target Fruit for Apple Harvesting Robot

In order to lessen picking time of apple harvesting robot,the fast tracing recognition method of target fruit for apple harvesting robot was researched. Firstly, the first collected image was segmented by the OTSU dynamic threshold segmentation method based on R-G color feature, and the picking target fruit was determined based on the principle of the nearest to image center. Next, the target fruit in the follow images were traced and recognized with the improved fast mean-residual normalized product correlation template matching algorithm while the region of image process idea was smaller frame-by-frame continuously according to the correlated information between the acquired images. At last, the comparative tests, which took into account of different threshold segmentation methods, different matching recognition by gray value, brightness and contrast, and the recognition time with the new and old methods, were done, and results from experiments indicated that the method used is viable, decreases by 36%.