基于模板匹配的苹果果实识别

APPLE FRUIT RECOGNITION BASED ON TEMPLATE MATCHING

为了实现果蔬收获机器人的果实识别和定位,同时为了消除由于果实大小和生长角度对识别的影响,以自然环境下的果园苹果作为识别对象,采用模板匹配方法,计算目标图与模板图的归一化互相关系数矩阵,通过在空域对模板图像做尺度变换和角度旋转进而对目标图像进行模板匹配,实现对果园苹果的果实识别。实验结果表明,该模板匹配算法能有效识别自然环境下的苹果果实,可以用于机器人视觉系统的识别任务。

Aiming at realising the recognition and location of fruit by picking robot as well as eliminating the influence of size and growth angle of fruit on recognition,a way of template matching is used to calculate the normalized cross-correlation matrices between the target image and the template image in which the apples in the orchard of natural environment are taken as the recognising objects.The fruit recognition of the apples in orchard is achieved by scaling and rotating the template images in spatial frequency domain and then conducting template matching on the object images.Experimental results show that the algorithm of template matching can detect apple fruits of natural environment effectively and is able to be used in recognition task of robot visual system.