Large scale image mosaic construction for agricultural applications

1. A method for generating improved image mosaics, the method comprising:

using a set of overlapping images of a landscape to determine three-dimensional locations for a plurality of features of the landscape;

based on the three-dimensional locations of the plurality of features, determining a three-dimensional plane of interest Random Sample Consensus without homography; and

warping the overlapping images onto the plane of interest to form a mosaic of images;

improving the mosaic of images by identifying an area of the mosaic captured by a plurality of images, determining a distortion metric for each of the plurality of images, and selecting the image with the best distortion metric as the image for the area, wherein the set of overlapping images of the landscape comprises a subset of a large set of overlapping images of the landscape and wherein at least one of the plurality of images that capture the area are part of the large set of overlapping images but are not part of the set of overlapping images.

2. The method of claim 1 wherein determining a three-dimensional plane of interest comprises using a plane fitting algorithm that identifies three-dimensional locations that lie within the three-dimensional plane and three-dimensional locations that are outliers relative to the three-dimensional plane.

3. The method of claim 1 wherein using a set of overlapping images to determine three-dimensional locations comprises using a Structure from Motion technique.

4. The method of claim 3 wherein the Structure from Motion technique further provides a camera pose for each image in the set of overlapping images.

5. The method of claim 1 further comprising determining a camera pose for an image that is part of the large set of overlapping images but not part of the set of overlapping images by determining a relative camera pose between the image and an image that was part of the set of overlapping images.

6. The method of claim 1 wherein determining a distortion metric comprises normalizing a number of pixels in the image dedicated to the area by a distortion in those pixels.

7. An agricultural image system comprising:

an unmanned aerial vehicle having mounted thereon a camera configured to capture a set of captured images of a geographical area;

a processor, forming a mosaic from the set of captured images through steps comprising:

selecting a subset of the set of captured images;

using the subset of images to identify geographical features of the geographical area;

selecting a plane based on the geographical features; and

warping the subset of images onto the selected plane instead of an image to form a mosaic.

8. The system of claim 7 wherein the processor performs further steps comprising:

dividing the mosaic into blocks; and

for each block, identifying images in the set of captured images that include one pixel for the block and selecting one of the identified images as the image for the block.

9. The system of claim 8 wherein selecting one of the identified images comprises determining a quality measure for each identified image and selecting the identified image with the best quality measure.

10. The system of claim 9 wherein the quality of measure comprises a distortion level of the pixels assigned to the block.

11. The system of claim 7 wherein selecting a plane based on the geographical features comprises fitting a plane to the geographical features such that at least one geographical feature is considered an outlier.

12. The system of claim 7 wherein identifying geographical features comprises using structure from motion to identify the geographical features and camera poses for each image in the subset of images.

13. The system of claim 7 wherein the set of captured images is captured at a relatively low altitude.

14. A method of constructing a mosaic comprising:

defining a geographic area by selecting initial images for a mosaic;

dividing the geographical area into block areas; and

for each block area, identifying a plurality of images that contain pixels covering the block area, evaluating each image in the plurality of images, and selecting one of the plurality of images as the image for the block area based on the evaluations to thereby form an improved mosaic wherein evaluating each image comprises identifying the pixels in the image that correspond to the block area, determining a displacement of each identified pixel based on the location of the pixel in the image, and using the displacements of the pixels to form a quality measure for the image relative to the block area.

15. The method of claim 14 wherein defining the geographical area comprises:

using the initial images to identify three-dimensional features on the geographical area;

using the three-dimensional features to identify a plane for the geographical area; and

rectifying the initial images onto the plane.

16. The method of claim 15 wherein identifying the plane comprises performing plane fitting such that at least one three-dimensional feature is an outlier to the plane.

17. The method of claim 16 wherein during the step of identifying three-dimensional features on the geographical area, poses for cameras of the initial images are determined.