Systems and methods for analyzing remote sensing imagery
Disclosed systems and methods relate to remote sensing, deep learning, and object detection. Some embodiments relate to machine learning for object detection, which includes, for example, identifying a class of pixel in a target image and generating a label image based on a parameter set. Other embodiments relate to machine learning for geometry extraction, which includes, for example, determining heights of one or more regions in a target image and determining a geometric object property in a target image. Yet other embodiments relate to machine learning for alignment, which includes, for example, aligning images via direct or indirect estimation of transformation parameters.
1. A method of determining a geometric object property in a target image, comprising:
receiving, at an extractor, training images;
receiving, at the extractor, a geometric object corresponding to a portion of one or more of the training images;
receiving, at the extractor, training geometric object properties, wherein each of the training geometric object properties identifies a corresponding geometric object;
receiving, at the extractor, first one or more parameters related to orientation of illumination source that associates formation of the training images for the training images;
creating, at the extractor, at least one of a classifier or a regression model configured to determine a geometric object property for an image based on the training images, the geometric object, the training geometric object properties, and the first one or more parameters;
receiving, at the extractor, a target image;
receiving, at the extractor, a target geometric object corresponding to a portion of the target image;
receiving, at the extractor, second one or more parameters related to orientation of illumination source that associates formation of the target image; and
determining, at the extractor using the at least one of a classifier or a regression model, a target geometric object property associated with the target geometric object.
2. The method of claim 1 , wherein each of the training images and the target image is one of a red-green-blue, panchromatic, infrared, ultraviolet, multi-spectral, or hyperspectral image.
3. The method of claim 1 , wherein at least one of the training geometric object properties is at least one of slope, pitch, dominant pitch, material, area, height, or volume, and wherein the target geometric object property is at least one of slope, pitch, dominant pitch, material, area, height, or volume.
4. The method of claim 1 , wherein the geometric object is at least one of a point, a contour, an area, or a binary mask, and wherein the target geometric object is at least one of a point, a contour, an area, or a binary mask.
5. The method of claim 1 , wherein the first one or more parameters comprises at least one of time, date, sun direction, sun position, latitude, longitude, or object material, and wherein the second one or more parameters comprise at least one of time, date, sun direction, sun position, latitude, longitude, or object material.