基于深度学习方法的复杂场景下车辆目标检测

Vehicle detection based on deep learning in complex scene

针对实际交通场景下的车辆目标,应用深度学习目标分类算法中具有代表性的Faster R-CNN框架,结合Image Net中的车辆数据集,把场景中的目标检测问题转换为目标的二分类问题,进行车辆目标的检测识别。相比传统机器学习目标检测算法,基于深度学习的目标检测算法在检测准确度和执行效率上优势明显。通过本实验结果分析表明,该方法在识别精度以及速度上均取得了显著的提高。

In recent years,deep learning algorithm has been widely used in the field of object detection.Vehicle objects come from the real traffic scene,this paper applied the faster R-CNN framework,which was a representative of the deep learning object classification algorithm,and combined with the ImageNet dataset,converted object detection problem into a binary classification problem in the scene to detection and recognization.Compared with target detection algorithm in traditional machine learning,it has obvious advantages in detection accuracy and execution efficiency based on deep learning.The experimental results show that the method has achieved a remarkable improvement in both recognition accuracy and speed.