A fast environmental change detection approach based on unsupervised multiscale texture clustering

In this paper, an analytic model for mapping land cover change that uses a novel multiresolution analysis in concert with image differencing to identify environmental change/no‐change areas is presented. The method adaptively chooses thresholds to segment targets from background, by using multiscale decomposition of the image difference. Changes due to atmospheric influences are automatically suppressed and resolution heterogeneity problems minimized. Examples that demonstrate the efficiency of the technique on medium spatial resolution Landsat multitemporal imagery are presented. Empirical evaluation supports the suitability of this technique for fast identification of changed scenes. As opposed to the traditional change detection methods, we directly detect land cover changes rather than simply pixels. This is evident from the less‐pixelated appearance of the change maps. The technique is recommended for regional/local environmental vulnerability and risk assessment as well for automated updating of GIS databases

Change detection · Climate change · Cluster analysis · Computer vision · Data mining · Environmental change · Geography · Identification (biology) · Image (mathematics) · Image resolution · Land cover · Land use · Pattern recognition (psychology) · Pixel · Remote sensing · Vulnerability (computing) · Artificial Intelligence · Computer Science · Geology · Remote Sensing and Land Use · Remote Sensing in Agriculture · Remote-Sensing Image Classification

• Textural Features for Image Classification

  Robert M Haralick, K Shanmugam et al.•IEEE Transactions on Systems,…•1973

  Add to list