灾情图中建筑物识别算法的研究

The Building Recognition Algorithm Research of the Disaster Images

破坏性地震会带来巨大的危害和损失,震后2小时内灾情信息极度匮乏.为了更加及时而准确地获取灾情图中的建筑物信息,本文提出了一种能够自动识别灾情图中建筑物轮廓的方法,通过对比多种图像边缘检测算法后,采用Canny算法获得灾情图中建筑物的边缘图像,利用Hough变换提取出直线,并根据直线间的位置关系生成直线关系图,进而检测识别出建筑物.实验结果表明:通过有针对性的图像建筑物特征提取,经过大量的训练后,计算机能自动识别图像中的建筑物灾情信息,并将其提取出来,该算法对部分损坏严重的建筑物识别效果不佳,后续将进一步对算法进行改进.该算法为后续实现在通信带宽受限的情况下,对灾情图像进行有针对性的处理提供帮助,具有重要意义.

Destructive earthquake will bring great harm and loss,and the disaster information is extremely scarce within 2 hours after the earthquake. In order to get the building information in disaster images more timely and accurately,a method of automatically identifying the outlines of buildings in disaster images is presented in this paper. After comparing various image edge detection algorithms,the Canny algorithm is used to obtain the edge image of the building in the disaster images,and the straight lines are extracted by Hough transformation,and the line diagram is generated according to the position relationships between the lines,and then the building is detected and identified. The experimental results show that after a lot of building feature extraction training from images,the computer can automatically identify and extract the disaster information of the building in the images. The algorithm is not effective in identifying partially serious damaged buildings. It should be improved in the future. It has great significance for the subsequent realization of the targeted treatment of the disaster images in the case of limited communication bandwidth.