Object measurement using deep learning analysis of synthetic aperture radar backscatter signatures
AssigneeOrbital Insight, Inc.
InventorsXin Li, Matthew Dunseth Wood
Life of this patent
Estimating the term…
Abstract
A system is configured to receive synthetic aperture radar (SAR) backscatter signatures of a geographical area including the object of interest from a SAR device. The system also extracts feature vectors from the SAR backscatter signature based on the intensity values of the SAR backscatter signature. The system inputs the one or more feature vectors into a neural network model. The system receives, as output from the neural network model, coordinate values indicating one or more visual features of the object of interest. Using these coordinate values, the system determines one or more measurements of the object of interest.
Claims19
A method for processing synthetic aperture radar (SAR) signatures from a SAR device, the method comprising:
- receiving an SAR backscatter image representing an SAR backscatter signature of a geographical area including an object of interest at a given time, the SAR backscatter signature including a two dimensional array of intensity values;
- modifying the SAR backscatter image by removing portions of the SAR backscatter image until the object of interest is within a threshold value of a center of the image;
- extracting one or more features from the modified SAR backscatter image based on intensity values of the modified SAR backscatter image;
- inputting the one or more features into a neural network model, the neural network model trained using training data, the training data including one or more training sets, each training set comprising labeled SAR backscatter signatures of objects of interest;
- receiving, as an output from the neural network model, coordinate values indicating one or more visual features of the object of interest;
- and
- determining one or more measurements of the object of interest based on the coordinate values, wherein each of the one or more measurements characterize a state of the object of interest at the given time.
A system, comprising:
- a synthetic aperture radar (SAR) device to capture an SAR backscatter image representing an SAR backscatter signature of a geographic area that includes an object of interest at a given time, the SAR back scatter signature including two dimensional array of intensity values;
- and
- an object measurement system configured to:
- receive the SAR backscatter image;
- modify the SAR backscatter image by removing portions of the SAR backscatter image until the object of interest is within a threshold value of a center of the image;
- extract one or more features from the modified SAR backscatter image based on the intensity values of the modified SAR backscatter image;
- input the one or more features into a neural network model, the neural network model trained using training data, the training data including one or more training sets, each training set comprising labeled SAR backscatter signatures of objects of interest;
- receive, as an output from the neural network model, coordinate values indicating one or more visual features of the object of interest;
- and
- determine one or more measurements of the object of interest based on the coordinate values, wherein each of the one or more measurements characterize a state of the object of interest at the given time.
A non-transitory computer readable storage medium, storing computer-readable instructions, that when executed by a processor of a computing system, causes the processor to:
- receive an SAR backscatter image representing an SAR backscatter signature of a geographical area including an object of interest at a given time, the SAR backscatter signature including a two dimensional array of intensity values;
- modify the SAR backscatter image by removing portions of the SAR backscatter image until the object of interest is within a threshold value of a center of the image;
- extract one or more features from the modified SAR backscatter image based on intensity values of the modified SAR backscatter image;
- input the one or more features into a neural network model, the neural network model trained using training data, the training data including one or more training sets, each training set comprising labeled SAR backscatter signatures of objects of interest;
- receive, as an output from the neural network model, coordinate values indicating one or more visual features of the object of interest;
- and
- determine one or more measurements of the object of interest based on the coordinate values, wherein each of the one or more measurements characterize a state of the object of interest at the given time.
Dependencies are inferred from the claim text.