Tomato Maturity Detection Method Based on Improved YOLOv5
The detection of tomato ripeness is of great significance and valuable for automated tomato harves-ting.To address the current issues of low detection,recognition accuracy and missed detections,a tomato ripe-ness detection method based on improved YOLOv5 is proposed.Firstly,SE attention module and BiFPN network are added to the original YOLOv5,enabling it to simultaneously focus on the features of small target objects in both channel and space,thereby enhancing the fusion ability of network features.Secondly,replacing the activa-tion function in the original network structure with the FReLU activation function can achieve pixel level spatial modeling ability,further improve the detection accuracy,and increase the robustness of the model.Experiments have shown that the improved YOLOv5 model has improved accuracy,recalland average accuracy by 4.8%,4.0%,and 3.0%respectively.Although the improved model has increased the number of parameters and com-putational complexity by 0.2M and 0.6G,it has improved the detection of tomatoes with different maturity levels and can provide technical support for automated picking.
maturityobject detectionattention mechanismFReLU activation function
NETL
NSTL
万方数据
