Cloud feature detection
AssigneeBAE SYSTEMS plc
InventorsRoderick Buchanan, James Duncan Revell
Life of this patent
Estimating the term…
Abstract
Disclosed are apparatus and a method for detecting cloud features. The method comprises: obtaining (402) image data (401); identifying (404) regions of the image data corresponding to a below-horizon region and an above-horizon region; classifying (404) one or more parts of the above-horizon region as a cloud feature; determining (409) an identification model specifying a visual appearance of some or all of the parts of the above-horizon region that have been classified as a cloud feature; and using the determined identification model, classifying (404), as cloud features, those regions of the of the below-horizon region whose visual appearance is as specified by the identification model.
Claims19
A method of detecting cloud features, the method comprising:
- obtaining image data;
- identifying regions of the image data corresponding to a below-horizon region and an above-horizon region;
- classifying one or more parts of the above-horizon region as sky;
- classifying, as a cloud feature, each part of the above-horizon region that has not been classified as sky;
- determining an identification model specifying a visual appearance of some or all of the parts of the above-horizon region that have been classified as a cloud feature;
- and
- using the determined identification model, classifying, as cloud features, those regions of the of the below-horizon region whose visual appearance is as specified by the identification model;
- wherein
- the step of classifying one or more parts of the above-horizon region as sky comprises:
- computing image gradients for Red-Green-Blue channels of the image data in directions tangential and normal to a horizon line;
- smoothing, by averaging over a local circular neighborhood, the computed image gradients for the Red-Green-Blue channels;
- summing the smoothed Red-Green-Blue channel image gradients to produce a combined smoothness value of each pixel in the directions tangential and normal to the horizon line;
- selecting a part of the above-horizon region in which, for each pixel in that part, the combined smoothness values of that pixel in directions tangential and normal to the horizon line are below a predetermined threshold value;
- and
- classifying the selected part as sky if properties of pixels within the selected part fulfil one or more predetermined criteria.
An apparatus for detecting cloud features comprising:
- one or more cameras configured to obtain image data;
- and
- one or more processors operatively coupled to the one or more cameras and configured to:
- identify regions of the image data corresponding to a below-horizon region and an above-horizon region;
- classify one or more parts of the above-horizon region as sky;
- classify, as a cloud feature, each part of the above-horizon region that has not been classified as sky;
- determine an identification model specifying a visual appearance of some or all of the parts of the above-horizon region that have been classified as a cloud feature;
- and
- using the determined identification model, classify, as cloud features, those regions of the of the below-horizon region whose visual appearance is as specified by the identification model;
- wherein
- the one or more processors are further configured to:
- compute image gradients for Red-Green-Blue channels of the image data in directions tangential and normal to a horizon line;
- smooth, by averaging over a local circular neighborhood, the computed image gradients for the Red-Green-Blue channels;
- sum the smoothed Red-Green-Blue channel image gradients to produce a combined smoothness value of each pixel in the directions tangential and normal to the horizon line;
- select a part of the above-horizon region in which, for each pixel in that part, the combined smoothness values that pixel in directions tangential and normal to the horizon line are below a predetermined threshold value;
- and
- classify the selected part as sky if properties of pixels within the selected part fulfil one or more predetermined criteria.
Dependencies are inferred from the claim text.