融入纹理特征的输电杆塔鸟巢图像检测方法

Bird’s Nest Image Detection Method for Transmission Towers Incorporating Texture Features

为实现航拍输电杆塔图像鸟巢自动检测,首先结合输电杆塔的特性提出了输电杆塔框架提取算法——选择合适颜色空间对图像进行背景粗分割,利用Canny边缘检测和Hough变换筛选出合适的连通域,把图像分成10×10像素大小的盒子并结合杆塔的几何特征提取输电杆塔框架。然后,在确定的杆塔区域内搜索出符合鸟巢样本HSV颜色分量的连通区域,作为候选鸟巢区域。最后,通过分析鸟巢样本纹理特征的灰度共生矩阵特征量,使用惯性矩特征量得到疑似鸟巢区域分类。利用现有输电杆塔图像进行鸟巢实例检测,检测结果表明该方法有效。

In order to realize automatic bird’s nest detection of aerial transmission pole tower images,firstly, a transmission pole tower frame extraction algorithm is proposed based on the characteristics of transmission poles and towers, which selects appropriate color space for coarse segmentation of image background, and uses Canny edge detection and Hough transform to screen out appropriate connected domain.The image is divided into 10×10 pixel boxes and the frame of transmission tower is extracted by combining the geometric features of the tower.Then, in the determined tower area, the connected region matching the HSV color component of nest sample is searched as the candidate nest region.Finally, the suspected bird’s nest region classification is obtained by analyzing the Gray-level co-occurrence matrix feature of bird’s nest sample texture feature and using the moment of inertia feature.The bird nest example is detected by using the existing transmission tower image, and the detection results show that the method is effective.