Detecting graph-based spatial outliers: algorithms and applications (a summary of results)
Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining
A Unified Approach to Detecting Spatial Outliers
Geoinformatica
| Hi-index | 0.00 |
As spatial outliers in remotely sensed imageries are regarded as abnormal values reflecting abnormal natural or man-made phenomena, there have been continuous researches to detect such outliers. On the other hand, in statistics, methods based on spatial autocorrelations are developed to detect the outliers. These ideas may be combined to detect forest fire pixels in the satellite imageries from NASA's AQUA platform. Reasoning comes from the fact that the forest fire detection means finding spatial outliers using spatial variations of brightness temperature. Thus, in this paper, we propose a new forest fire detection approach with Moran scatterplot analysis that is based on spatial outlier detection methods. This approach is a bit revised one than the one already proposed by the same researchers last year, which is more effective in detecting the forest fire. The proposed approach was tested to evaluate its effectiveness. The evaluation was done by comparing the results with the MODIS fire product provided by the NASA MODIS Science Team, and the results of last year's research. The evaluation results showed the revised approach has a good potential in detecting the fire pixels.