基于YOLOv7的轻量级低照度目标检测算法

Lightweight Low-Light Object Detection Algorithm Based on YOLOv7

低照度目标检测是目标检测任务中常见的挑战之一。通用的目标检测方法在低照度条件下性能会明显下降,而现有的低照度目标检测方法会造成大量的计算资源消耗,并不适合部署在计算能力受限的设备上。为应对上述问题,提出一种端到端的轻量级目标检测算法LL-YOLO。针对低照度图像中特征信息不明显、难以学习与辨识的问题,设计低照度图像生成算法,通过生成低照度图像来训练检测器,帮助其学习低照度环境下的特征信息;并对检测器网络结构进行调整,减少特征信息在计算过程中的损失,提高模型对特征信息的敏感度。针对低照度图像中特征信息受噪声影响严重的问题,提出聚合周边信息的A-ELAN模块,使用深度可分离卷积与注意力机制捕获周边信息,增强获得的特征信息,减弱噪声的影响。实验结果表明,LL-YOLO算法在低照度目标检测数据集ExDark上平均精度均值(mAP@0.5)达到81.1%,相较直接训练的YOLOv7-tiny算法提高11.9百分点,相比于其他算法具有较强竞争力。

Lowlight object detection is a major challenge in object detection tasks.Conventional methods for object detection exhibit significant performance degradation under lowlight conditions,and existing lowlight object detection methods consume excessive computational resources,making them unsuitable for deployment on devices with limited computing capabilities.To address these issues,this study proposes an endtoend lightweight object detection algorithm called lowlight YOLO(LLYOLO).To tackle the problem of unclear and difficulttolearn features in lowlight images,a lowlight image generation algorithm is designed to generate lowlight images for training the detector,assisting it in learning feature information in lowlight environments.In addition,the network structure of the detector is adjusted to reduce the loss of feature information during computation,thereby enhancing the model’s sensitivity to feature information.Furthermore,to mitigate the problem of severe noise interference on feature information in lowlight images,an aggregation ELAN(AELAN)module for aggregating peripheral information is proposed that uses depthwise separable convolution and attention mechanisms to capture contextual information,enhance the obtained feature information,and weaken the impact of noise.Experimental results demonstrate that the LLYOLO algorithm achieves a mAP@0.5 of 81.1%on the lowlight object detection dataset ExDark,which is an improvement of 11.9 percentage points over that of the directly trained YOLOv7-tiny algorithm.The LLYOLO algorithm exhibits strong competitiveness against existing algorithms.