面向无人船的天空区域检测算法
Sky Area Detection Algorithm for USV
-
摘要: 无人船视觉图像受波浪和倒影的影响,水面区域颜色和纹理复杂多变,增加了无人船的视觉感知难度。不同于传统方法处理整个图像,本文研究主要围绕特征相对稳定的天空区域检测任务。根据图像的边缘信息,以及水上场景不同区域的饱和度和亮度分布情况,提出一种基于多信息融合的天空检测算法。首先,为了去除噪声和局部纹理,保留图像不同区域的结构信息,提高算法中区域分割效果,采用变分法对图像进行预处理。然后采用基于图的视觉分割算法,利用边界信息将图像分割成不同区域。其次,对水上视觉场景进行分析,利用在HSV颜色空间下不同区域在图像上的分布特征,初步估计天空区域位置。最后,利用分割结果,准确确定天空区域。实验结果表明,在不同场景下,所提出算法具有较高精度和稳定的性能。Abstract: A sky detection algorithm of multi-information fusion type is developed. The detection of sky area is accomplished by checking the edge information of image and the saturation and luminance distribution feature of image parts. The image signal is preprocessed with the variational method to remove the noise and detail texture in order to see the structure of image parts clearly. The processed image is segmentate according to the edge information by means of the visual image segmentation algorithm. The color distribution in the HSV color space in image sections is analyzed to pick up the sky area. Final sky detection is accomplished after further examining.
下载: