雾霾图像增强算法对目标检测性能影响研究

针对雾霾环境下目标检测率低,容易造成交通事故等问题,提出了基于图像增强的动态双阈值算法。该算法针对雾霾天气下,传统的检测算法目标检测率低、虚警率高等问题,利用大气散射模型及联合双边滤波算法首先对原始雾霾图像的增强处理,然后再进行目标检测。分别使用动态双阈值、基于均值滤波的动态双阈值、基于直方图均衡化的动态双阈值、基于拉普拉斯算子的动态双阈值目标检测算法对不同程度雾霾环境下的实拍车辆运动视频进行目标检测,并用11097组图像数据对比分析改进算法的最佳检测率、最差检测率、平均检测率和虚警率。实验结果表明,改进算法目标检测率高、虚警率低,有助于减少交通事故,更加适用于雾霾环境下图像目标检测。

Impacts analysis of object detection algorithm considering image enhancement in haze environment

Considering traffic accidents prone to the cause of lower object detection rate in haze condition, a dynamic dual-threshold object detection algorithm based on image enhancement was investigated in this paper. To cope with the problem of burr and discontinuity caused by the traditional edge detection operator in the severe haze days, an atmospheric scattering model based image enhancing method was put forward to improve the quality of the original image. Then object detection method of dynamic threshold, dynamic threshold based on mean-value filtering, dynamic dual threshold based on histogram equalization and dynamic dual threshold based on Laplace operator were employed to process and analyze the video of vehicle motion under variable haze conditions, and the application scopes and corresponding thresholds of these dynamic dual thresholds were determined. Finally, the best detection rate, the most poor detection rate, the average detection rate and false alarm rate of the improved algorithm are compared with 11097 images. The result illustrates that the improved algorithm is more suitable for object detection in haze condition and can help to reduce the traffic accident.