An Assisted Change Detection System

NASA has a critical need for intelligent change detection algorithms andgeoprocessing techniques for digital imagery archives of earth sciencedata obtained from satellites and aerial platforms. Currently, the changedetection process is slow, labor-intensive, and costly. Change detectionmethods that rely on pixel differencing detect all changes between imageryand are thus not well suited for detecting target-specific changes, suchas riparian zones within a selected watershed. Thus, targets must firstbe extracted from both sets of imagery before target changes can bedetected. Visual Learning Systems (VLS) proposes to design and demonstratethe feasibility of an intelligent target-specific change-detection modulebased on VLS's novel, proprietary machine-learning methods. VLS's approach is expected to provide a mechanism for detecting object-specificchanges that is 50-200 times faster than existing techniques whilemaintaining the same level of accuracy.