Multispectral anomaly detection

Inventors

Burge, Mark J. • CHENEY, Jordan

Assignees

Noblis Inc

Abstract

Techniques for detecting anomalies in multispectral image data, and more specifically for detecting presentation attacks by using multispectral image data in biometric security applications, are provided. In some embodiments, a system may receive multispectral image data and generate an estimation of a first image of a plurality of images of the multispectral image data, wherein the estimation is based on other images of the multispectral image data, but not the first image itself. The estimation may then be compared to the first image to generate an indication as to whether the multispectral image data represents a presentation attack. In some embodiments, a system may receive multispectral training image data and may extract features from the data to generate and store a network architecture for predicting relationships of multispectral images of subjects.

Core Innovation

The invention relates to multispectral anomaly detection for biometric presentation attack detection using cross-spectral analysis. Multispectral image data representing a subject at different wavelength ranges is aligned into a spectral cube and processed to extract cross-wavelength features. Feature extraction is performed using convolutional neural networks and a Gabor wavelet filter bank, including hierarchical part-based representations via Hierarchical Part-based TensorFaces (HPBT) and tensor dictionaries.

The invention models expected inter-wavelength relationships and generates an estimated image for a withheld wavelength using an auto-encoder and a CNN-based “hallucinated” image generation approach. The generated estimation is produced based on a subset of the plurality of images not containing the first image.

The invention compares the first image to the generated estimation across wavelength ranges and produces a presentation-attack indication using similarity metrics, including Siamese networks and an ensemble of CNNs. Optional decision logic supports thresholding, metric comparisons, Bayesian equations/framework, and spatial heat-map localization to generate a spatial indication.

Claims Coverage

The document provides three independent claim sets that share the same core workflow: receive multispectral image data at different wavelength ranges, generate an estimation of a withheld image using the remaining images, compare the real image to the estimation, and generate an indication of whether the data represents a presentation attack. No additional inventive feature refinements are explicitly included in the independent claim text beyond this workflow for these independent claims.

Across the independent claim sets, the inventive coverage centers on multispectral presentation-attack detection via cross-spectral image estimation for a withheld image, followed by comparing the withheld image to its estimation and producing an indication based on the comparison.

Stated Advantages

Documented Applications