Improved agricultural mosaic and 3D model reconstruction
Large scale mosaics are constructed by stitching together images obtained from unmanned aerial vehicles (UAVs) flying at low altitudes. This novel method generates improved image mosaics by using overlapping images of a landscape after determining a three-dimensional plane of interest and warping the overlapping images onto that plane to form a mosaic. The process begins by dividing a geographical area into block areas and analyzing several images of each area. Next, the algorithm selects the best image for each location and then independently rectifies each image to obtain the corresponding portion of the mosaic. The novel technique avoids costly joint-optimization over the entire sequence and is validated using challenging input sequences motivated by agricultural applications. The method has recently been extended for 3D point cloud reconstruction and semantic feature extraction.
MN-IP Try and Buy |
---|
Try
|
Buy
|
Unique algorithm reduces computational burden
Precision agriculture currently uses image analysis that is inaccurate or requires too much computing power/time to be fast enough for farmers. Stitching together numerous homogeneous landscape images is a challenge. Existing image stitching/mosaicking methods rely on inter-image homographic computation based on a planar scene assumption for images taken from high altitudes. These methods also avoid scale and resolution changes by relying on primarily translational motion at fixed altitudes. This new approach overcomes these limitations and performs well even with aerial images taken from low altitudes by a UAV performing complex motions. The algorithm creates ground geometry from a small sample of images, uses the highest quality images to match images to the ground geometry and achieves an accurate field map without the computationally intensive process of stitching images together. It can accommodate a large number of images because the joint optimization required for robustly estimating the world structure is performed on a very small fraction of the images, yet all images in the input sequence are used and a novel method chooses the best image for each ground patch.
Phase of Development
- Prototype developed
Benefits
- Provides useful information for agricultural managers
- Faster than existing stitching methods
- Less computationally intensive
Features
- Accounts for dynamic movement and altitude variability
- Enables navigation of ground based robots/drones through crop rows
- Can use aerial images taken from low altitudes
Applications
- Corn and soybean crops
- Precision agriculture
Publications
Interested in Licensing? |
---|
The University relies on industry partners to scale up technologies to large enough production capacity for commercial purposes. The license is available for this technology and would be for the sale, manufacture or use of products claimed by the issued patents. Please contact us to share your business needs and technical interest in this technology and if you are interested in licensing the technology for further research and development. |