基于改进YOLOv8n的货车盲区目标检测方法

Object Detection in the Blind Spot of Truck Based on Improved YOLOv8n

摘要:

针对货车盲区范围大，背景复杂，目标尺度变化大，现有的货车盲区检测方法效果差，容易漏识别等问题，提出了改进的YOLOv8n货车盲区目标检测算法。在主干网络中加入混合局部通道注意力模块（MLCA），提升网络局部空间特征提取能力；改进特征融合网络，由尺度序列特征融合模块（SSFF）融合图像多个尺度的深层语义信息，三重特征编码模块（TFE）捕获目标的局部细节；最后采用Inner-CIoU作为边框损失函数，提高边框检测精度。实验结果表明，在自建的车辆行人数据集上，相比传统YOLOv8n算法，所提算法的平均检测精度提升了3.14%，在货车盲区目标检测中具有更优的检测能力。

Abstract:

Aiming at the problems of large scope of truck blind zone, complex background, large variation of target scale, poor effect of existing truck blind spot detection methods and easy to miss recognition, an improved YOLOv8n truck blind zone target detection algorithm is proposed. A mixed local channel attention module(MLCA) is added to the backbone network to improve the local spatial feature extraction capability of the network; the feature fusion network is improved by the scale sequence feature fusion module(SSFF) to fuse the deep semantic information of multiple scales of the image, and the triple feature encoding module(TFE) to capture the local details of the target; and finally, Inner-CIoU is adopted as the bezel loss function to improve the border detection accuracy. The experimental results show that on the self-built vehicle-pedestrian dataset, the proposed algorithm improves the average detection accuracy by 3.14% compared with the traditional YOLOv8n algorithm, and it has better detection capability in target detection in the blind zone of trucks.