【摘要】 针对现有目标检测算法对于目标遮挡以及小目标造成的误检漏检等问题，以YOLOv7作为基础框架，提出了一种复杂驾驶环境下目标检测的改进方法。首先在原模型上引入CBAM注意力机制，抑制背景干扰信息；使用NWD作为算法的损失函数，增强对小目标的检测能力。实验结果表明，在公开数据集KITTI上，相较原算法准确率提高了4.5%，有效提升了驾驶环境中目标检测的精度。更多还原

【Abstract】 In order to solve the problems of false detection and missed detection caused by the existing object detection algorithms, an improved object detection method in complex driving environment is proposed based on YOLOv7. Firstly,CBAM attention mechanism is introduced into the original model to suppress background interference information. Secondly,NWD is used as the loss function of the algorithm to enhance the detection ability of small object. Experimental results show that, on the public data set KITTI, the accuracy rate is improved by 4.5% compared with the original algorithm.更多还原