基于颜色空间变量的输电线图像分类及特征提取

Image classification and feature extraction of transmission line based on color space variable

针对输电线图像背景复杂多变和单一图像处理方法难以有效处理各种背景类型输电线图像的问题,提出一种基于颜色空间变量的输电线图像分类及特征提取方法。首先根据输电线图像各颜色空间的变量值与图像特征之间的关系对图像进行分类。然后根据不同类别图像特征选用适合的滤波方法通过两次滤波结合去噪,并采用自适应直方图分段均衡化增强图像对比度。通过对Otsu算法得出的阈值进行线性变换确定canny边缘检测参数,提取输电线路边缘。最后根据输电线形状特征和概率霍夫直线变换与形态学运算提出一种边缘优化方法,较好地去除非输电线边缘。结果表明:该方法可以有效处理各种背景类型图像,为输电线路图像智能化处理提供了一种新的思路和方法。

Considering the complicated and variable background of transmission line image, the single method of image processing is difficult to effectively deal with various background types of transmission line images, thus a method of image classification and feature extraction based on color space variable is proposed. Firstly, according to the relationship between the variable value of each color space and the image feature of the transmission line image, the image is classified. Then, the appropriate filtering method is selected according to the different image features. The noise is removed by combining two filtering, and the adaptive histogram segmentation equalization is used to enhance the image contrast. The parameters of the canny edge detection are determined by linear transformation of the threshold obtained by the Otsu algorithm to extract transmission line edge. Finally, according to the shape feature of transmission line and the probability Hough linear transformation and morphological operation, a method of edge optimization is proposed and the non-edge of transmission line is removed better. The results show that the method can effectively deal with various background types of images, providing a new idea and method for the intelligent processing of transmission line images.