激光与光电子学进展, 2017, 54 (5): 051501, 网络出版: 2017-05-03
交通场景静态显著性区域检测 下载: 759次
Static Saliency Region Detection in Traffic Scenes
机器视觉 显著性区域 奇异值分解 交通场景 强光抑制 machine vision saliency region singular value decomposition traffic scene strong light suppression
摘要
交通场景的显著目标检测能够为自动决策或辅助驾驶系统提供重要信息。基于视觉的底层特性, 提出了一种基于亮度空间和对立颜色空间的多特征空间奇异值分解的交通场景显著性区域快速检测方法, 为交通标志检测和场景语义分析提供有效信息。首先, 在亮度空间中, 利用奇异值分解确定强光区域并进行强光抑制, 检测抑制强光后的亮度特征显著性区域; 其次, 在双对立颜色空间中保留部分奇异值重构的区域作为颜色特征显著性区域; 最后, 对各个特征空间的显著性区域进行线性相加并将相加结果中的显著性区域作为交通场景目标检测的候选区域。实验结果表明, 算法在复杂光照和背景下具有较好的稳健性。
Abstract
The traffic scene saliency detection can provide important information for the automatic decision or driving assist system. Based on the underlying visual characteristics, a new method for the rapid saliency detection in traffic scene is proposed, which is based on the singular value decomposition of multi-feature space (the color space and the opposition color space). First of all, in the brightness space, using the singular value decomposition to estimate the high light region and suppress the high light area to detect the brightness saliency. Secondly, in the dual opposition color space, the region of partial singular value reconstruction is retained as the color feature saliency region. Finally, the saliency region of each feature space is added together and the saliency region is the candidate region of the target detection of traffic scene. Experimental results show that the proposed method has good robustness in illumination variation and complex background scenes.
方志明, 崔荣一, 金璟璇. 交通场景静态显著性区域检测[J]. 激光与光电子学进展, 2017, 54(5): 051501. Fang Zhiming, Cui Rongyi, Jin Jingxuan. Static Saliency Region Detection in Traffic Scenes[J]. Laser & Optoelectronics Progress, 2017, 54(5): 051501.