【摘要】 针对草莓病害图像在检测时存在背景复杂、目标小导致难以被精确检测的问题，本文提出一种基于多元感受野与EResPANet的草莓病害检测算法.首先，该算法使用多元感受野特征标定网络替换YOLOv7-Tiny的主干网络，抑制冗余信息，解决主干网络特征逐层提取时小目标病害丢失问题；最后，通过设计EResPANet网络，避免网络在深层特征提取时，目标信息被复杂背景干扰而导致无法检测的问题.实验结果表明，本文提出的方法相比YOLOv7-Tiny算法在mAP上提高了10.3%,证明本文算法可实现草莓各类病害的准确检测.更多还原

【Abstract】 Aiming at the problem that strawberry disease images are difficult to be accurately detected due to the complex background and small targets during detection, this paper proposes a strawberry disease detection algorithm based on multivariate receptive field and EResPANet.First, the algorithm uses the multivariate receptive field feature calibration network to replace the backbone network of YOLOv7-Tiny, suppresses the redundant information, and solves the problem of small target disease loss during the layer-by-layer extraction of the features of the backbone network; finally, through the design of the EResPANet network, it avoids the problem that the target information of the network is interfered with by the complex background during the deep feature extraction, which leads to the problem of non-detection.The experimental results show that the method proposed in this paper improves 10.3% in mAP compared with the standard YOLOv7-Tiny algorithm, which proves that the algorithm in this paper can realize the accurate detection of various types of diseases in strawberry.更多还原