基于改进的YOLOv5龙眼果柄位置识别与检测方法

Improved YOLOv5 Longan Fruit Stalk Position Recognition and Detection Method Based on YOLOv5

提出一种基于改进的YOLOv5s目标检测算法的龙眼果柄识别方法,以满足高效果柄位置识别及加工作业精细化管理需求。针对龙眼体积小,果柄与果实颜色相近等影响因素,引入CBAM注意力机制,强化对果柄位置特征的关注学习。同时简化模型复杂度,减少检测时间。在颈部网络中采用加权双向特征金字塔网络BIFPN,优化多尺度特征融合机制,提升模型检测精度。实验结果表明,改进的YOLOv5算法检测识别效果优于初始模型,能满足自动化龙眼剥壳中对果柄位姿检测精度及效率要求。

A longan fruit stalk recognition method based on the improved YOLOv5s target detection algorithm was proposed in order to meet the demand for high effect stalk position recognition and refined management of processing operations.Aiming at the influence factors such as the small size of longan and the similarity between the stalk and the colour of the fruit,the CBAM attention mechanism is introduced to strengthen the attention learning of the stalk location features and weaken the influence of the rest of the information on the recognition results,At the same time,the model complexity is simplified to reduce the detection time.The weighted bidirectional feature pyramid network BIFPN is used in the neck network to optimise the multi-scale feature fusion mechanism and improve the model detection accuracy.The experimental results show that the improved YOLOv5 algorithm detects and recognises better than the initial model,and can meet the requirements for the precision and efficiency of fruit stalk position detection in automated longan shelling.