基于深度学习的红外图像检测技术优化

Optimization of Infrared Image Detection Technology Based on Deep Learning

红外目标检测具有全天候、可探测距离远、不受大气和光照条件影响等优点,特别适用于无人驾驶汽车在夜间弱光条件环境,可实现较好的检测效果,具有广阔的应用前景。但红外图像存在分辨率低、边缘模糊、对比度差的问题,导致检测精度下降。因此,以Flir数据集为基础,首先利用线性空间域滤波技术实现红外图像增强,提高图像边缘清晰度,并利用部署了BiFPN特征网络的YOLOv5算法训练图像。结果表明,增强后的红外图像训练模型精确度提升12.1%,召回率提升6.7%,平均精确度0.5提升20.7%,平均精确度0.5:0.95提升4.9%,证明该研究可有效提升红外图像的检测精度。

Infrared target detection has the advantages of all-weather,long detectable distance,and not affected by atmospheric and lighting conditions.It is particularly suitable for unmanned vehicles in low light conditions at night and can achieve good detection results,with broad application prospects.However,infrared images suffer from low resolution,blurred edges,and poor contrast,leading to a decrease in detection accuracy.Therefore,based on the Flir dataset,the first step is to use linear spatial domain filtering technology to enhance infrared images,improve image edge clarity,and train images using the YOLOv5 algorithm deployed with BiFPN feature network.The results showed that the accuracy of the enhanced infrared image training model increased by 12.1%,the recall rate increased by 6.7%,the average accuracy increased by 20.7% at 0.5,and the average accuracy increased by 4.9% at 0.5:0.95,proving that this study can effectively improve the detection accuracy of infrared images.