Computer Engineering and Applications ›› 2019, Vol. 55 ›› Issue (6): 178-185.DOI: 10.3778/j.issn.1002-8331.1712-0107

Previous Articles     Next Articles

Skyline Detection Algorithm Based on Multiple Feature Extraction Fusing Edge Correction

TU Bing1,2,3 , ZHANG Xiaofei1,3, PAN Jianwu1,3, ZHANG Guoyun1,2,3, ZHOU Zixuan1,3   

  1. 1.School of Information and Communication Engineering, Hunan Institute of Science and Technology, Yueyang, Hunan 414006, China
    2.Key Laboratory of Optimization and Control for Complex Systems, College of Hunan Province, Hunan Institute of Science and Technology, Yueyang, Hunan 414006, China
    3.Laboratory of Intelligent-Image Information Processing, Hunan Institute of Science and Technology, Yueyang, Hunan 414006, China
  • Online:2019-03-15 Published:2019-03-14


涂  兵1,2,3,张晓飞1,3,潘建武1,3,张国云1,2,3,周紫璇1,3   

  1. 1.湖南理工学院 信息与通信工程学院,湖南 岳阳 414006
    2.湖南理工学院 复杂系统优化与控制湖南省普通高等学校重点实验室,湖南 岳阳 414006
    3.湖南理工学院 IIP创新实验室,湖南 岳阳 414006

Abstract: Focused on the issue of high robustness and high accuracy detection of skyline, a skyline detection algorithm based on multiple feature extraction and edge correction is proposed. The multi-eigenvalues of the training pixels randomly in sky and non-sky regions are extracted according to texture information and color information. Then, the multi-eigenvalues are used to train a classifier based on Support Vector Machine(SVM) to obtain the initial position coordinates of skyline. Next, the Canny operator method is used to detect the edge of the gray image. And the linear five neighborhood search algorithm is used to correct the position of the initial coordinate, finally skyline coordinates of original images are obtained. The proposed algorithm is tested on the Web Set and the Basalt Hills Set, the results indicate that the proposed method can effectively detect the skyline coordinates, reduce the interference of other pixels to some extent and make the skyline more smoothly.

Key words: skyline detection, Support Vector Machine(SVM), Canny edge detection, linear five-neighborhood search algorithm

摘要: 针对天际线的高鲁棒性与高准确率检测问题,提出了一种多特征提取与边缘校正融合的天际线检测算法。采用Gabor纹理特征和颜色特征提取天空与非天空区域随机训练像素点的多特征值,接着采用支持向量机(Support Vector Machine,SVM)对多特征值训练得到分类器,从而检测出天际线的初始坐标位置;接着采用Canny算子对灰度化图像进行边缘检测,并利用线性五邻域搜索算法对初始坐标位置进行校正,最终得到天际线坐标位置。最后将所提算法在Web数据集和Basalt Hills数据集上进行测试,实验结果表明:提出的算法能有效地检测出较复杂图像场景中的天际线位置,在一定程度上减少了图像中相关像素点的干扰,使检测出的天际线更加平滑。

关键词: 天际线检测, 支持向量机, Canny边缘检测, 线性五邻域搜索算法