Abstract: The event of tripping operation caused by bird’s nest in transmission line becomes a serious threat to the national grid. An automatic detection method of bird’s nest is proposed. For the area of tower in image composed of wire rod in different directions, the density of different directions line segments in per portioned block is regarded as the judgment for identifying the area of tower in UAV image. The detected subblock region is clustered, and then the tower area is identified. The connected region matching the HSV of the sample nest is searched and selected as the candidate of the nest area. And then characteristic parameters of shape and characteristic quantities of the intensity variance of roughness and characteristic quantities of the inertia matrix of texture in candidates are analyzed. Tested images from UAV verify that the method can identify nests effectively from tower with the interference of background removed.

Key words: UAV inspection, nest detection, tower identification, color and texture blending

摘要： 由于鸟巢造成的输电线路跳闸事件频频发生，已严重威胁到国家电网的安全运行，为了降低复杂背景的影响，提出了一种自动检测铁塔上鸟巢的方法，首先识别巡检图像上铁塔所在区域，考虑到铁塔是由不同方向的线材构成的空间图像，将巡检图像分块并分析不同方向的线段密度，判决是否属于铁塔区域，将检测的分块铁塔区域聚类，进而识别铁塔区域。在铁塔区域内，搜索符合鸟巢样本的HSV颜色特征量的连通区域，作为候选的鸟巢区域，分析候选鸟巢区域的形状特征参数，描述鸟巢粗糙度的灰度方差特征量，描述鸟巢纹理的惯性矩特征量，通过对无人机巡检采集的输电线路图像的测试，验证了这种方法能有效排除背景的干扰，有效检测出铁塔上的鸟巢。

关键词:
无人机巡检, 鸟巢检测, 铁塔识别, 颜色与纹理融合