基于改进分水岭算法的苹果识别方法

Apple Recognition Method Based on Improved Watershed Algorithm

目的针对自然生长环境下重叠苹果的识别问题,提出一种基于距离变换机制的改进分水岭算法。方法首先,通过对比实验优选出苹果在RGB、HSI和Lab颜色空间的图像处理模型;然后,采用距离变换和归一化处理方法对苹果的位置特征进行内部标记;进而基于锐化方法提高原图像的边缘强度,并结合内部标记符改进分水岭算法,实现苹果轮廓的分离;最后,基于最小外接圆法提取苹果的单连通区域,实现重叠苹果的识别。结果简化了目标与背景的分割过程,通过内部标记法改进了分水岭算法,实现重叠目标的轮廓重建,解决重叠苹果的过度分割问题。结论实验结果表明,该算法对重叠苹果的准确识别率高于92%,验证了所提方法的有效性。

Considering the recognition problem of overlapping apples in natural growing environment,an improved watershed algorithm based on distance transformation mechanism was proposed.Firstly,the image processing models of apple in RGB,HSI and Lab color space are optimized through comparative experiments.Then,the distance transformation and normalization methods were used to evaluate the internal label of apple’s position characteristics.Then,the edge intensity of the original image was improved based on the sharpening method,and the watershed algorithm was improved based on the internal label to realize the separation of apple contours.Finally,the simply connected region of the apple was extracted based on the minimum enclosing circle method to realize the recognition of overlapping apples.The proposed method simplifies the segmentation process of target and background,improves the watershed algorithm through internal labelling method,realizes the contour reconstruction of overlapping targets,and solves the problem of over-segmentation of overlapping apples.The experimental results show that the accuracy of the algorithm is higher than 92%,which invalidates the effectiveness of the proposed method.