As an alternative for combining individual detectors, the present data fusion system combine information in DSM, TIR sensor data, Hyperspectral and infrared colour images through a high-level data fusion system that uses Bayesian statistics involving weights-of evidence modelling (WoE). To determine the efficacy of the system, an analysis of change-detection was performed. The high-level data fusion system is capable of detecting changes in man-made features automatically where there is little prior information. Multiclass segmented images were obtained from the data captured by four airborne remote sensing sensors. The results of the Bayesian method are accurate as the weights are based on statistical analysis. Changes in features such as colour of roofs, parking areas, open land areas, newly built structures, and the presence or absence of vehicles are extracted automatically in a densely populated city in Japan.
Airborne Digital Remote Sensing Data:
High-resolution airborne imagery of two dates from ADS40, AISA, VNIR-Eagle and TABI sensors are used to test the system. DSM data at GSD of 50cm is obtained from the ADS40. All these sensors’ data are registered automatically at the time of acquisition by GPS-IMU technology to experiment the effectiveness of the data fusion system.
Our modelling system detects changed areas for a period time based on the distribution of known occurrences (e.g. change-detected pixels). Notable advantage of this automatic fusion system are missing data can be accommodated, patterns with complex spatial geometry can be modelled robustly, and the system can be implemented in any Image processing and GIS software environments as an add-on module.
The system is scalable and functional to any applications. Example Fields are Medical Imaging where new locations of Tumor Diagnosis and Mineral Deposits Explorations study. From a set of input images, our fusion system produces a single image that has more complete information. With improved reliability and capability the system is useful for human or machine perception.
The authors thank PASCO-Japan for data and for encouraging this research work.
Biography of Authors:
Babu Madhavan obtained PhD (1995) in Remote Sensing (SAR) from IIT-Bombay. Later, completed two post-doc research projects at Keio University, Japan and joined PASCO in 2002 and served as CEO for its office in India (2005 to 2010). With 24 years work experience in image processing and computer vision research, Babu published more than 85 papers. Recently Babu won ‘Geospatial World Innovation Award’ for 3D modelling technology development in The Netherlands for his research at Softopia Japan.
Tadashi Sasakawa, Vice-President of PASCO, obtained PhD.Eng (2005) from Hokkaido University. He joined PASCO in 1982 and remained engaged in dissemination of GIS. In 2000, he was instrumental in the introduction of aircraft sensors ADS40, UCD, LIDAR, and automation of digital photogrammetry. In 2005, he established Satellite Business Division for commercialization and applications of TerraSAR-X imageries.