基于改进YOLOv8的工厂行人检测算法

Factory pedestrian detection algorithm based on improved YOLOv8

针对工厂中行人检测算法精度不足,存在误检、漏检等问题,提出一种基于改进YOLOv8的工厂行人检测算法。首先,在YOLOv8的C2f模块中引入卷积块注意力机制模块(CBAM),以帮助主干网络聚焦于关键特征并抑制非关键特征,从而提升模型对遮挡物和小目标的检测准确度;其次,在Neck网络中将卷积神经网络Conv模块替换成CoordConv模块,以充分利用该模块的定位能力,从而解决目标检测中的定位准确性问题,提升模型对空间位置的感知能力;最后,采用Inner-IoU损失函数替代原始的CIoU损失函数,来提高目标检测边界框的回归精度。在自制的工厂行人图像数据集(3 600张图像)上进行了训练和测试,实验结果表明:相较于基础YOLOv8算法,改进YOLOv8算法在平均精度均值(mAP)和每秒帧率(FPS)方面分别提高了2.26%和35.6 f/s,验证了改进算法在检测性能上的提升。

A factory pedestrian detection algorithm based on improved YOLOv8 is proposed to address the issues of insufficient accuracy,false positives,and missed detections in pedestrian detection algorithms in factories.The convolutional block attention mechanism(CBAM)module was introduced into the C2f module of YOLOv8 to help the backbone network focus on key features and suppress non key features,thereby improving the model's detection accuracy for occlusions and small targets.The convolutional neural network Conv module is replaced by the CoordConv module in the Neck network to make full use of the positioning ability of the module,so as to solving the positioning accuracy in object detection and improve the model's perception of spatial position.The Inner IoU loss function is used to replace the original CIoU loss function to improve the regression accuracy of object detection bounding boxes.A self-made pedestrian image dataset in a factory(3600 images)are trained and tested.The experimental results show that in comparison with the basic YOLOv8 algorithm,the improved YOLOv8 algorithm can improve the average accuracy of the average mAP(mean average precision)and the frame rate FPS(frame rate per second)by 2.26%and 35.6 f/s,respectively,which can verify the improvement of the detection performance of the improved algorithm.