基于图像结构信息的城市房屋震害特征自动提取技术
|
摘要
由于常规边界检测算法难以完全满足城市房屋震害特征提取的需求 ,本研究提出了一种基于最优方法的边界跟踪方法 ,并利用ENVIIDL二次开发语言实现了基于区域结构和纹理统计特性相结合的损坏房屋自动识别算法。最后以 1976年我国唐山地震的黑白航空影像为例 ,进行倒塌房屋的自动识别 ,得到了较为满意的结果。结果表明 ,利用本文所提出的震害建筑自动识别方法 ,在提取震害房屋信息方面是有效的 ,其结果与目视结果大体一致。在完全倒塌的试验区 ,震害房屋的识别结果与目视解译的结果几乎完全一致 ,符合程度在 99%左右 ;在其部分倒塌区 ,自动识别的结果与目视解译的结果符合程度也达到了 89%。在这项自动识别理论中 ,由于采用了基于区域的图像处理与分析技术 ,所以保持了房屋建筑最基本的特性———区域特征 ,完全不同于基于光谱特征的震害房屋识别技术 ,故而可以得到较好的自动识别结果。
A new method is suggested to detect and classify damaged buildings automatically based on statistical information of texture and structure features of damaged buildings' imagery regions. The software system that is designed following the theory and technique of IDL (interactive data language) has been development. Based on the previous analysis and, selecting black and white aerial photograph of the 1976 Tangshan earthquake as example, the method of automatic pattern recognition is tested and classification of damaged buildings is proposed. Comparisons between the results of automatic recognition and classification and the manual interpretation show that the precision of automatic pattern recognition is about 85%. By introducing the techniques of image processing and the imagery regions analysis, the fundamental features (regional special properties) of the buildings can be kept. Therefore better results of automatic recognition can be obtained.
A new method is suggested to detect and classify damaged buildings automatically based on statistical information of texture and structure features of damaged buildings' imagery regions. The software system that is designed following the theory and technique of IDL (interactive data language) has been development. Based on the previous analysis and, selecting black and white aerial photograph of the 1976 Tangshan earthquake as example, the method of automatic pattern recognition is tested and classification of damaged buildings is proposed. Comparisons between the results of automatic recognition and classification and the manual interpretation show that the precision of automatic pattern recognition is about 85%. By introducing the techniques of image processing and the imagery regions analysis, the fundamental features (regional special properties) of the buildings can be kept. Therefore better results of automatic recognition can be obtained.
引文
1 陈鑫连,魏成阶编.地震灾害的航空遥感信快速评估与救灾决策[M ].科学出版社,1995.
2 Y .Yusuf,M .Matsuoka,F .Yamazaki.Detectionofbuildingdamagesduetothe2001Gujarat,Indiaearthquakeusingsatelliteremotesensing[A].Proceedingofthe22ndAsianConferenceonRemoteSensing
3 Matsuoka,M .,andF .Yamazaki.IdentificationofDamagedAreasduetothe1995Hyogoken-NanbuEarthquakeusingSatelliteOpticalImages[A].Proceedingsofthe19thAsianConferenceonRemoteSensing,Q9,1998,1～6
2 Y .Yusuf,M .Matsuoka,F .Yamazaki.Detectionofbuildingdamagesduetothe2001Gujarat,Indiaearthquakeusingsatelliteremotesensing[A].Proceedingofthe22ndAsianConferenceonRemoteSensing
3 Matsuoka,M .,andF .Yamazaki.IdentificationofDamagedAreasduetothe1995Hyogoken-NanbuEarthquakeusingSatelliteOpticalImages[A].Proceedingsofthe19thAsianConferenceonRemoteSensing,Q9,1998,1～6
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心