摘要
设计了一个基于YOLACT++深度学习算法的槟榔检测模型。针对传送带上采集的槟榔图片分割精度低和预测框不精确造成槟榔分级准确率低的问题。在模型主干网络中引入改进Res2Net模块,改善槟榔掩模分割精度。在模型边界框回归损失中引入CIoU损失函数,提高预测框的检测精度。结果表明,改进模型的掩模mAP相较YOLACT++、Mask R-CNN、SOLOv2分别高出5.20%,4.09%,2.37%。预测框mAP相较YOLACT++、Mask R-CNN分别高出5.41%,4.90%。相较于模型改进前分级准确率提升2.12%。
A betel nut detection model based on YOLACT++deep learning algorithm is designed,aiming at the problem of low accuracy of betel nut grading caused by low segmentation accuracy and inaccurate prediction frame of the betel nut images collected on the delivery belt.The improved Res2 Net module is introduced into the model backbone network to improve the accuracy of the betel nut mask segmentation.The CIoU loss function is introduced into the model bounding box regression to improve the detection accuracy of the prediction box.The results show that the mask mAP of the improved model is 5.20%,4.09%,and 2.37% higher than that of YOLACT++,Mask R-CNN and SOLOv2,respectively.Compared with YOLACT++and Mask R-CNN,the prediction frame mAP is 5.41%and 4.90% higher respectively.Compared with the model improvement before the classification accuracy rate is increased by 2.12%.
作者
舒军
王祥
舒心怡
SHU Jun;WANG Xiang;SHU Xinyi(School of Electrical and Electronic Engineering,Hubei Univ.of Tech.,Wuhan 430068,China;Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System,Wuhan 430058,China;Wuhan Britain-China School,Wuhan 430068,China)
出处
《湖北工业大学学报》
2022年第4期29-35,共7页
Journal of Hubei University of Technology