Integrating GNSS, IMU, and Imagery for Automatic Orthomosaic Generation
The use of orthomosaic images from aerial or satellite data are increasingly common. While current acquisition methods are cost-effective on a national or regional scale, local scale imagery is prohibitively expensive for many target applications. In this paper we present a combined hardware and software solution, developed at the Geospatial Research Centre, which aims to reduce the cost of acquiring and processing imagery and related data in order to produce orthomosaics in a cost-effective manner on a small, local scale. The hardware component consists of a combined GNSS and inertial solution for determining the position and orientation of a sensor, typically a consumer-grade camera such as a digital SLR. The combination of imagery and navigation metadata allows images to be directly geo-referenced by projecting them on to readily available surface models. Refinements to this initial processing are also presented, which account for boresight and lens calibration error; automatically establishing a correspondence between image features for bundle adjustment; and reducing the visual appearance of any residual misalignments in the final mosaic. The use of commodity sensors and automated processing is an important step in reducing the cost of image acquisition and orthomosaic generation. The methods described are illustrated using two sample sequences. The first is a set of visible images captured from a digital SLR, and the second a set of frames extracted from a thermal video sequence. These two sequences demonstrate the range of imagery that can be processed, which can support applications ranging from environmental monitoring and precision agriculture to urban planning and infrastructure maintenance.