Combining Pixel- and Object-Based Machine Learning for Identification of Water-Body Types From Urban High-Resolution Remote-Sensing Imagery

Abstract

Water is one of the vital components for the ecological environment, which plays an important role in human survival and socioeconomic development. Water resources in urban areas are gradually decreasing due to the rapid urbanization, especially in developing countries. Therefore, the precise extraction and automatic identification of water bodies are of great significance and urgently required for urban planning. It should be noted that although some studies have been reported regarding the water-area extraction, to our knowledge, few papers concern the identification of urban water types (e.g., rivers, lakes, canals, and ponds). In this paper, a novel two-level machine-learning framework is proposed for identifying the water types from urban high-resolution remote-sensing images. The framework consists of two interpretation levels: 1) water bodies are extracted at the pixel level, where the water/shadow/vegetation indexes are considered and 2) water types are further identified at the object level, where a set of geometrical and textural features are used. Both levels employ machine learning for the image interpretation. The proposed framework is validated using the GeoEye-1 and WorldView-2 images, over two mega cities in China, i.e., Wuhan and Shenzhen, respectively. The experimental results show that the proposed method achieved satisfactory accuracies for both water extraction [95.4% (Shenzhen), 96.2% (Wuhan)], and water type classification [94.1% (Shenzhen), 95.9% (Wuhan)] in complex urban areas.

Cloud Extraction from Chinese High Resolution Satellite Imagery by Probabilistic Latent Semantic Analysis and Object-Based Machine Learning

“…A series of state-of-the-art algorithms recently were developed for remote-sensing image classification, such as SVM, Extreme Learning Machine (ELM), Decision Tree, Random Forest (RF) and Tree Bagger (TB). Huang, et al (2015) experimentally verified that SVM outperformed other machine-learning methods in target extraction from remote sensing images. In addition, RBF-SVM demonstrated the best stability for non-linear classification [31].…”

“…Huang, et al (2015) experimentally verified that SVM outperformed other machine-learning methods in target extraction from remote sensing images. In addition, RBF-SVM demonstrated the best stability for non-linear classification [31]. In this paper, the result is a non-linear classification model that can be used to classify new test superpixel.…”

Water-Body types identification in urban areas from radarsat-2 fully polarimetric SAR data

“…In urban areas, many water resources are now facing threats from nutrient enrichment, organic, and inorganic pollution (Palmer et al, 2015). Besides, with the rapid urban expansion and population growth, water resources in urban areas are also gradually decreasing (Niemczynowicz, 2009;Huang et al, 2015). To improve the understanding of physical, chemical, and biological properties of water resources, and to monitor the illegal use and pollution, the demands for precise and real-time water monitoring in large areas are increasing.…”

“…of water resources rather than their areas. Different water-body types show totally different functions and have different effects on the urban ecology and environments (Huang et al, 2015). Rivers play an important role in providing domestic water, regulating climate, and transportation; lakes in urban areas are mainly used for conserving water and irrigation; ponds provide places for fishery and planting aquatic products.…”

“…Huang et al (2015) presented some promising results about water-body types identification from high-resolution optical images using shape features and textures. He et al (2014) proposed a feature extraction method and introduced to very high-resolution TerraSAR image for classification, in their experiments, rivers and ponds are analyzed along with five other land cover types, the training and validation samples are at the patch level (64 × 64) which are artifically selected from the original TerraSAR image.…”