基于轻量化YOLO-v3的绿熟期番茄检测方法

Detection of mature green tomato based on lightweight YOLO-v3

准确识别定位绿熟期番茄果实是实现其自动采摘的必要前提。由于绿熟期番茄的表面颜色仍为青色与叶片、枝干颜色接近,特别是存在叶片、枝干遮挡和果实重叠类型的图像,传统的图像检测处理方法不能准确进行定位。为解决此问题,采用改进的深度学习目标检测算法YOLO-v3进行番茄检测,将原算法的骨干网络DarkNet-53改为更轻量化的Mobilenet-v1。结果表明:轻量化YOLO-v3算法将模型大小缩小为原来的39.38%,训练速度提高3.88倍,验证集的平均精度均值达到98.69%,测试集的平均精度均值达到98.28%。所采用的轻量化YOLO-v3检测算法可实现对绿熟期番茄的实时目标检测,更适合在移动设备和嵌入式端进行部署,为更加高效的番茄自动采摘奠定基础。

The accurate detection of mature green tomato is a prerequisite of automatic picking.Owing to color similarity between the surface color of mature green tomato,leaf,and branch,as well as the existence of images with leaf,branch occlusion and fruit overlapping,the detection performance of traditional image detection was bad.To solve this problem,an improved YOLO-v3 algorithm was designed for mature green tomato image detection,where the original DarkNet-53 backbone was replaced by the lightweight Mobilenet-v1.Compared with the original YOLO-v3 algorithm,the lightweight algorithm reduced the model size by 39.38%,increased the train speed by 3.88 times,and the mean average precision of the validation set and test set was 98.69%and 98.28%,respectively.Therefore,the proposed lightweight YOLO-v3 could successfully realize real-time detection of mature green tomato and is more suitable to be implement in mobile and embedded devices,which will make automatic picking of tomato more efficient.