Moving Car Detection and Spectral Restoration in a Single Satellite WorldView-2 Imagery

A novel approach to automatic detection of moving vehicles from a single satellite WorldView-2 imagery is presented. The technique is based on the time gap between three sensor band groups: panchromatic, multispectral, and four new additional multispectral bands. The entire process is automatic and includes movement estimation followed by moving object spectral restoration and construction of spatially built object profiles to estimate the movement direction and velocity. The approach neither relies on external information like road data or site models, nor is limited to vehicle type. The performance of the new approach is demonstrated via detection of several vehicle types on WorldView-2 satellite imagery of the San Francisco area.