A framework for registering UAV-based imagery for crop-tracking in Precision Agriculture

Multiple types of images provide useful information about a crop, but image fusion is still a challenge in Precision Agriculture (PA). We describe a framework which manages a multi-layer registration model of heterogeneous images obtained by an unmanned aerial vehicle (UAV) by proposing pair-to-pair steps through a registration method invariant to intensity differences, allowing us to connect different aerial images with significant differences. Correction of deformed images is treated as a first step to end up with our registration algorithms. These methods conform the base of more advanced systems that combine 2D and spatial information, therefore it represents the link of several types of images. The evaluation shows the flexibility of our framework when dealing with different requirements. Effectiveness of the Enhanced Correlation Coefficient method is proved and thus shown as a suitable method for theregistration of heterogeneous images.

Automated summary

This paper introduces a framework that successfully connects different types of aerial images through multi-layer registration using a registration method that is invariant to intensity differences. Correction of deformed images as the first step of the registration algorithm lays the foundation for more advanced systems.