面向大型建筑物形变监测的图像角点检测方法

针对大型建筑物在实际灾害中因各种致变因素的作用发生形变，当超出一定限度时会演变成灾难的实际问题进行研究，通过引入高斯核卷积函数、初始像素点集的邻域作差筛选以及利用最小核值相似区的思想来筛选角点提出一种对大型建筑物进行形变监测的改进算法，并与四种相应算法进行了对比实验。结果表明：本文改进算法相较于其他算法在建筑物图像形变监测中正确率平均提升了近15%，平均检测时间缩短了近21%。可见本文改进算法提升了大型建筑物形变监测中的多尺度检测能力、减少了计算数据量、提升了角点检测准确性。

Image Corner Detection Method for Large Structure Deformation Analysis

Large buildings would be deformed due to the various mutable factors in actual disasters. When a certain limit is exceeded, it can turn into a disaster. In order to solve this problem, the related research was carried out in this paper. The Gaussian kernel convolution function is introduced, the initial pixel point set is screened by the neighborhood difference, and the corner points is filtered through the idea of the minimum kernel value similarity region. Based on the above three points, an improved algorithm for deformation monitoring of large buildings is proposed and compared with the four corresponding algorithms. The experimental results show that the average correct rate of the improved algorithm in building image deformation monitoring is increased by nearly 15%, and the average detection time is reduced by nearly 21%. It is concluded that the multi-scale detection capability has been improved, the amount of computational data has been reduced, and the accuracy of corner detection has been improved by adopting the improved algorithm in this paper.