基于重叠边缘的夜间重叠番茄识别

Recognition of overlapping tomatoes based on overlapping edges at night

为更准确地识别出夜间重叠番茄,提出一种基于重叠边缘检测的夜间重叠番茄识别算法.该算法采用基于OTSU法进行图像分割,利用Canny算子进行边缘检测,并从中识别出重叠边缘,进一步基于重叠边缘应用一种距离就近法判断重叠番茄前后位置关系,最后基于圆拟合对重叠番茄中前未被遮挡番茄进行识别.试验结果表明:无枝叶遮挡时,前后位置关系判断正确率为96%,前未被遮挡番茄识别正确率为91.7%;当有枝叶遮挡时,前后位置关系判断正确率为90%;当被遮挡率低于50%时,前未被遮挡番茄识别正确率为83.3%;该算法可实现夜间双果重叠番茄前后位置关系判断及前未被遮挡番茄识别.

To realize the recognition of overlapping tomatoes at night,a recognition algorithm for overlapping tomatoes at night was designed based on overlapping edge detection.First,overlapping tomato images captured at night were segmented using the OTSU method.Second,edges were extracted using Canny operator,and overlapping edges were recognized.Third,the minimum distance method was used to judge the anteroposterior position of tomatoes in overlapping regions.Finally,the front tomato without occlusion in overlapping regions was recognized based on circle fitting.Test results show that the accuracy rate is 96%for the recognition of tomatoes’anteroposterior position in overlapping regions,while the accuracy rate is 91.7%for the recognition of the front tomato in overlapping regions under conditions that overlapping tomatoes arent’occluded by leaves and branches.The accuracy rate is 90%for anteroposterior position recognition under condition that overlapping tomatoes are occluded by leaves and branches,while the accuracy rate is 83.3%for the front tomato recognition under condition that the occlusion rate is less than 50%.The algorithm can realize tomatoes’anteroposterior position judgment and the front tomato recognition in overlapping regions at night.