US 12,657,680 B1Granted

Multi-modal automotive paint authentication and repainting detection system

AssigneeUVeye Ltd.

InventorAmir Hever

Life of this patent

Estimating the term…

Loading patent…

Abstract

There is provided a method of automatically detecting paint anomalies on a vehicle, comprising: operating illumination elements configured for generating illuminations of different types at first different frequency bands and/or at first different polarization angles, positioned for illuminating the vehicle, operating image sensors for capturing the images, wherein the image sensors are of different modalities configured for sensing at second different frequency bands and/or at second different polarization angles configured for capturing images of a vehicle, extracting features from the images, generating a multi-dimensional dataset by aggregating the features, feeding the multi-dimensional dataset into a machine learning model, obtaining from the machine learning model, an indication of a region of the vehicle with a paint anomaly, generating an enhanced image depicting the region of the vehicle with the paint anomaly overlaid with a visual indication of the region with the paint anomaly, and presenting the enhanced image on a display.

Claims15

Dependent claims

A system for automatically detecting paint anomalies on a surface of a vehicle, comprising:

a plurality of illumination elements configured for generating a plurality of illuminations of different types at a first plurality of different frequency bands and/or at a first plurality of different polarization angles, positioned for illuminating the vehicle;
a plurality of image sensors of a plurality of different modalities configured for sensing at a second plurality of different frequency bands and/or at a second plurality of different polarization angles configured for capturing a plurality of images of a vehicle;
a data interface configured to access and/or receive a plurality of images captured by the plurality of image sensors during illumination by the plurality of illumination elements;
at least one processor configured for:
operating the plurality of illumination elements for illuminating the vehicle;
operating the plurality of image sensors for capturing the plurality of images;
extracting a plurality of features from the plurality of images;
generating a multi-dimensional dataset by aggregating the plurality of features;
feeding the multi-dimensional dataset into a machine learning model;
obtaining from the machine learning model, an indication of a region of the vehicle with a paint anomaly;
generating an enhanced image depicting the region of the vehicle with the paint anomaly overlaid with a visual indication of the region with the paint anomaly;
and
presenting the enhanced image on a display,
wherein the at least one processor is configured for creating a training dataset of a plurality of records by:
creating a record for each sample vehicle of a plurality of sample vehicles including a first subset with a paint anomaly and a second subset without a paint anomaly:
operating the plurality of illumination elements for illuminating the sample vehicle;
operating the plurality of image sensors for capturing a plurality of images of different modalities of the sample vehicle;
extracting at least one feature from each of the plurality of images of different modalities;
and
generating a multi-dimensional dataset from the plurality of features,
wherein the record for the sample vehicle includes the multi-dimensional dataset and a ground truth label indicating whether the sample vehicle depicted the paint anomaly or did not depict the paint anomaly;
and
training the machine learning model on the training dataset.

20 more elements

A system for automatically detecting paint anomalies on a surface of a vehicle, comprising:

a plurality of illumination elements configured for generating a plurality of illuminations of different types at a first plurality of different frequency bands and/or at a first plurality of different polarization angles, positioned for illuminating the vehicle;
a plurality of image sensors of a plurality of different modalities configured for sensing at a second plurality of different frequency bands and/or at a second plurality of different polarization angles configured for capturing a plurality of images of a vehicle;
a data interface configured to access and/or receive a plurality of images captured by the plurality of image sensors during illumination by the plurality of illumination elements;
at least one processor configured for:
operating the plurality of illumination elements for illuminating the vehicle;
operating the plurality of image sensors for capturing the plurality of images;
extracting a plurality of features from the plurality of images;
generating a multi-dimensional dataset by aggregating the plurality of features;
feeding the multi-dimensional dataset into a machine learning model;
obtaining from the machine learning model, an indication of a region of the vehicle with a paint anomaly;
generating an enhanced image depicting the region of the vehicle with the paint anomaly overlaid with a visual indication of the region with the paint anomaly;
and
presenting the enhanced image on a display,
wherein the plurality of illumination elements and the plurality of image sensors are arranged into the plurality of imaging modalities, selected from:
(i) an ultraviolet (UV) fluorescence imaging subsystem including at least one UV illumination element operating at about 365 nanometer (nm) wavelength, at least one fluorescence camera equipped with at least one filter for isolating specific wavelength ranges selected from 400-700 nm emitted by the paint in response to the UV illumination,
(ii) a polarized light analysis subsystem including polarized light sources with adjustable polarization angles, and image sensors associated with polarization filter configured to analyze reflected light characteristics,
(iii) a spectroscopic examination subsystem including near-infrared spectroscopy devices covering wavelengths from about 800-2500 nm, and
(iv) and a red-green-blue (RGB) camera.

15 more elements

Dependencies are inferred from the claim text.