Morphotrak, LLC

Idemia Identity & Security USA LLC

Other

Peter Zhen-Ping Lo

Hui Chen

Darrell Hougen

Hui Zhu

Stephen R Gregg

Jianguo Wang

Laurent Semy Kazdaghli

Thomas Gruschus

Jean-Christophe Francois Marie Fondeur

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving mission specific data. Converting the mission specific data into attribute classes recognizable by image recognition classifiers, where the image recognition classifiers are pre-trained to detect objects corresponding to the attributed classes within digital images. Obtaining, from a wearable camera system, low resolution images of a scene and high resolution images of the scene. Detecting, within the low resolution images using a low resolution image classifier, an object that corresponds to one of the attribute classes. In response to detecting the object that corresponds to one of the attribute classes, providing, for presentation to a user of the wearable camera system by a presentation system, an first alert indicating a potential detection of the suspect, and providing the high resolution images to a high resolution image classifier. Obtaining, from the high resolution image classifier, a confirmation of the detected object from the low resolution images. In response to obtaining the confirmation, providing, for presentation to the user by the presentation system, a second alert indicating confirmation that the object has been detected.

Classifying camera images to generate alerts

In some implementations, a method includes: receiving, from the camera, a sample image that includes a fingerprint and a mensuration reference device, where the sample image is associated with a resolution; identifying (i) a plurality of edge candidate groups within the sample image, and (ii) a set of regularity characteristics associated with each of the plurality of edge candidate groups; determining that the associated set of regularity characteristics indicates the mensuration reference device; identifying a ruler candidate group, from each of the plurality of edge candidate groups, based at least on determining that the associated set of regularity characteristics indicates the mensuration reference device; computing a scale associated with the sample image based at least on extracting a set of ruler marks from the identified ruler candidate group; and generating, based at least on the scale associated with the sample image, a scaled image.

Automatic ruler detection

A method for detection of altered fingerprints includes receiving, by at least one processor, an image of a fingerprint from a fingerprint reader. The image has an image resolution. The processor determines a spatial location of the fingerprint within the image. The processor crops the image around the spatial location to provide a cropped image. The processor generates multiple derived images using the cropped image, such that each derived image has the image resolution. The processor generates a multiple-channel image using the derived images. The processor scales the multiple-channel image to an image size. The processor generates a score using a machine learning model. The score is based on the multiple-channel image and is indicative of a likelihood that the fingerprint has been altered.

Method and system for detection of altered fingerprints

A list of possible mated minutiae in a reference fingerprint and a search fingerprint is identified. The list of possible mated minutiae is divided into different groups of mated minutiae using pose grouping based on rigid transformation parameters. For each group, the possible mated minutiae are filtered to get a set of consistent mated minutiae pairs having similar rigid transformation. The topological consistency between groups is determined. If any two groups are topologically consistent to each other, they are merged to form a (virtual) larger group. A non-linear transformation parameters are estimated from the mated minutiae in the merged group. The overall similarity of mated and non-mated minutiae in the overlapping area is determined after non-linear alignment of the two fingerprints, and the maximum similarity is output as the final similarity score between two fingerprints.

Distorted fingerprint matching using pose and minutia grouping

In some implementations, a method may be used for matching palm print images. A search template that identifies at least a plurality of distinctive search orientation blocks within a search orientation field may be initially generated for a search palm print image. A reference template may be obtained. A mated distinctive reference orientation block may be identified for each of the distinctive search orientation blocks. One or more regions of the search orientation field that include the distinctive search orientation blocks may be compared against one or more corresponding regions of the reference orientation field. An orientation similarity score between the search palm print image and the reference palm print image may be computed based on the comparison. A match may finally be determined if the computed orientation score satisfies a predetermined threshold value.

Palm print image matching techniques

Systems and methods may be used by an automatic fingerprint identification system to estimate patterns of a latent fingerprint. A latent fingerprint image, and a plurality of reference ridge flow maps may initially be obtained. Each reference ridge flow map may be associated with a particular fingerprint pattern. A latent ridge flow map for the obtained latent fingerprint image may be computed. One or more characteristics associated with the latent ridge flow map may be compared to one or more characteristics associated with each of the plurality of reference ridge flow maps. A similarity score between the latent ridge flow map and a particular reference ridge flow map may be computed for each of the plurality of reference ridge flow maps. One or more fingerprint patterns present within the latent fingerprint may then be determined and provided for output.

Latent fingerprint pattern estimation

In some implementations, systems and techniques are capable of performing a multi-stage tattoo matching operation to improve matching accuracy for various tattoo images&#x2014;namely small tattoo images, distorted tattoo images, low-quality tattoo images. The multi-stage tattoo matching operation described throughout includes two matching stages that compare a search tattoo image against a set of reference tattoo images. In the first matching stage, the system can use scale invariant feature transform (SIFT) descriptors, or other similar types of local descriptors, to identify a list of best-matched reference tattoo images from among the set of reference tattoo images. In the second matching stage, the system uses enhanced tattoo matching techniques to verify the non-matched points and determine whether their associated areas are matched or not.

Multi-stage tattoo matching techniques

In some implementations, an image recognition system is capable of using an iterative multi-stage image matching procedure to improve the accuracy and/or reliability of matching a search image with a small number of distinctive feature points to reference images. In the first stage, the system compares a search template for the search image to reference templates of each image within the set of reference images. The system determines a best-matched reference image from among the reference images of the set of reference images. In the second stage, the system compares the hybrid search template to one or more reference templates of reference images that is not the best-matched reference image. The comparison can be used to either compute match scores for reference images corresponding to the one or more reference templates.

Multi-stage image matching techniques

Some implementations of the systems and techniques described include a tattoo recognition system that is capable of improving the accuracy and efficiency associated with matching edge curves extracted from tattoo images, which are typically performed during a tattoo matching operation. The system can perform a matching operation in two stages&#x2014;a feature extraction stage and a matching stage. In the first stage, the system extracts one or more edge curves from the tattoo image. In the second stage, the system performs matching using a two-step comparison that compares features of the extracted edge curves.

Fast curve matching for tattoo recognition and identification

A list of possible mated minutiae in a reference fingerprint and a search fingerprint is identified. The list of possible mated minutiae is divided into different groups of mated minutiae using pose grouping based on rigid transformation parameters. For each group, the possible mated minutiae are filtered to get a set of consistent mated minutiae pairs having similar rigid transformation. The topological consistency between groups is determined. If any two groups are topologically consistent to each other, they are merged to form a (virtual) larger group. A non-linear transformation parameters are estimated from the the mated minutiae in the merged group. The overall similarity of mated and non-mated minutiae in the overlapping area is determined after non-linear alignment of the two fingerprints, and the maximum similarity is output as the final similarity score between two fingerprints.

A fingerprint matching system for fingerprint matching using virtual minutiae includes a fingerprint scanning device, including (i) an interface to a fingerprint scanner, and (ii) a sensor associated with the fingerprint scanner, configured to identify a match between a search fingerprint representing a fingerprint of a subject and a reference fingerprint from among a plurality of reference fingerprints representing fingerprints of a plurality of historical subjects; a database containing the plurality of reference fingerprints representing fingerprints of the plurality of historical subjects; and a computing device in communication with said database and said fingerprint scanning device.

Fingerprint matching using virtual minutiae

Systems and methods may be used by an automatic fingerprint identification system to estimate ridge flow maps in latent fingerprints. A latent fingerprint image and a plurality of reference ridge flow maps may initially be obtained. A latent ridge flow map for the obtained latent fingerprint image may be computed. One or more characteristics associated with the latent ridge flow map may be compared to one or more characteristics associated with each of the plurality of reference ridge flow maps. A similarity score between the latent ridge flow map and a particular reference ridge flow map may be computed for each of the reference ridge flow maps. The top closely matched ridge flow maps are used to improve the latent ridge flow map to enhance the latent fingerprint image to extract better set of minutia points to improve latent to ten-print matching accuracy.

Latent fingerprint ridge flow map improvement

In some implementations, a computer-implemented method includes: obtaining a captured image including, (i) a latent fingerprint to be captured, and (ii) a template, surrounding the latent fingerprint, that contains plurality of symbols; extracting, using an image processing technique, one or more characteristics for each of the plurality of symbols; calculating a resolution for the captured image based at least on the extracted one or more characteristics for each of the plurality of symbols; generating a reconstructed desired image, based at least on (i) the calculated resolution of the captured image, and (ii) the one or more extracted characteristics for each of the plurality of symbols; and providing the reconstructed desired image to a fingerprint identification system to perform a particular fingerprint operation.

Growing community of inventors

Mission Viejo, CA, United States of America

Peter Zhen-Ping Lo