The present disclosure involves systems, software, and computer implemented methods for transaction auditing. One example method includes determining valid pixel-based pattern(s) that are included in valid reference images. Fraudulent pixel-based pattern(s) that are included in fraudulent reference images are determined. A request to classify an image is received. A determination is made as to whether pixel values in the image match a valid pixel-based pattern or a fraudulent pixel-based pattern. In response to determining that the pixel values match a valid pixel-based pattern, a likelihood of classifying the first image as a valid image is increased. In response to determining that the pixel values match a fraudulent pixel-based pattern, a likelihood that the image as a fraudulent image is increased. The image is classified in response to the request as either a valid image or a fraudulent image based on the likelihoods.

The present disclosure involves systems, software, and computer implemented methods for transaction auditing. One example method includes determining valid pixel-based pattern(s) that are included in valid reference images. Fraudulent pixel-based pattern(s) that are included in fraudulent reference images are determined. A request to classify an image is received. A determination is made as to whether pixel values in the image match a valid pixel-based pattern or a fraudulent pixel-based pattern. In response to determining that the pixel values match a valid pixel-based pattern, a likelihood of classifying the first image as a valid image is increased. In response to determining that the pixel values match a fraudulent pixel-based pattern, a likelihood that the image as a fraudulent image is increased. The image is classified in response to the request as either a valid image or a fraudulent image based on the likelihoods.

In one aspect, a computerized method for anti-counterfeiting solution using a machine learning (ML) model includes the step of providing a pre-defined set of feature detection rules, a pre-defined set of edge detection rules, a pre-defined threshold percentage, an original seal, an original fingerprint of the original seal, and a pre-trained fingerprint identification model. The pre-trained fingerprint identification model is trained by a specified ML algorithm using one or more digital images of the original seal. With a digital camera of a scanning device, the method scans a seal whose authenticity is to be determined. The seal is used to secure a transportation container. The method uses the pre-defined set of feature detection rules to detect and extract an extracted feature image at a specified position on the seal. The method breaks down the extracted feature image of the seal into a ‘kn’ number of sub-images by forming a ‘k’ rows x ‘n’ columns of a grid of the extracted feature image. The method implements the pre-defined set of edge detection rules to extract an edge structure of at least one object in each of the ‘kn’ number of sub-images. The method generates a set of unique fingerprints by specified steps. The method includes generating a unique fingerprint corresponding to a unique number or a feature based on each extracted edge structure. For the set of unique fingerprints, the method generates a match percentage for the set of unique fingerprints using the pre-trained fingerprint identification model. The match percentage corresponds to a matching proportion between each unique fingerprint generated for the seal being verified and the original fingerprint of the original seal on which the pre-trained fingerprint identification model is trained.

Geometric extraction is performed on an unstructured document by recognizing textual blocks on at least a portion of a page of the unstructured document, generating bounding boxes that surround and correspond to the textual blocks, determining search paths having coordinates of two endpoints and connecting at least two bounding boxes, and generating a graph representation of the at least a portion of the page, the graph representation including the plurality of textual blocks, the coordinates of the vertices of each bounding box and the coordinates of the two endpoints of each search path.

Techniques are described that exclude use of “stop-fingerprints” from media database formation and search query to an automatic content recognition (ACR) systems based on media content fingerprints updated by stop-fingerprint analysis. A classification process is presented which takes in fingerprints from reference media files as an input and produces a modified set of fingerprints as an output by applying a novel stop-fingerprint classification algorithm. Architecture for the distributed stop-fingerprint generation is presented. Various cases, as stop-fingerprints generation for the entire reference database, stop-fingerprints generation for the individual reference fingerprint files, and temporal fingerprint classification obtained through intermediate steps of the temporal fingerprint classification algorithm are presented. A hash-based signature classification algorithm is also described.

Methods of processing video are presented to generate signatures for motion segmented regions over two or more frames. Two frames are differenced using an adaptive threshold to generate a two-frame difference image. The adaptive threshold is based on a motion histogram analysis which may vary according to motion history data. Also, a count of pixels is determined in image regions of the motion adapted two-frame difference image which identifies when the count is not within a threshold range to modify the motion adaptive threshold. A motion history image is created from the two-frame difference image. The motion history image is segmented to generate one or more motion segmented regions and a descriptor and a signature are generated for a selected motion segmented region.

The present disclosure involves systems, software, and computer implemented methods for transaction auditing. One example method includes determining valid pixel-based pattern(s) that are included in valid reference images. Fraudulent pixel-based pattern(s) that are included in fraudulent reference images are determined. A request to classify an image is received. A determination is made as to whether pixel values in the image match a valid pixel-based pattern or a fraudulent pixel-based pattern. In response to determining that the pixel values match a valid pixel-based pattern, a likelihood of classifying the first image as a valid image is increased. In response to determining that the pixel values match a fraudulent pixel-based pattern, a likelihood that the image as a fraudulent image is increased. The image is classified in response to the request as either a valid image or a fraudulent image based on the likelihoods.

Audio distortion compensation methods to improve accuracy and efficiency of audio content identification are described. The method is also applicable to speech recognition. Methods to detect the interference from speakers and sources, and distortion to audio from environment and devices are discussed. Additional methods to detect distortion to the content after performing search and correlation are illustrated. The causes of actual distortion at each client are measured and registered and learnt to generate rules for determining likely distortion and interference sources. The learnt rules are applied at the client, and likely distortions that are detected are compensated or heavily distorted sections are ignored at audio level or signature and feature level based on compute resources available. Further methods to subtract the likely distortions in the query at both audio level and after processing at signature and feature level are described.

An electronic device and method are disclosed herein. The electronic device includes: a processor, and a memory operatively connected to the processor. The processor implements the method, including: receive an input of handwritten text including characters, set corpus line key points indicating a substantial-top positions of the characters included in the handwritten text and base line key points indicating a substantial bottom positions of the characters, calculate at least a first feature value based on at least one corpus line key point from among the corpus line key points, and a second feature value based on at least one extracted base line key point from among the base line key points, respectively, input the calculated first feature value into a first neural network to cause the first neural network to generate a first result value, input the calculated second feature value into a second neural network to cause the second neural network to generate a second result value, and input the first result value and the second result value into a fully connected neural network to generate a third result value.

An image processing apparatus includes a character determining unit configured to divide the read image into multiple blocks, each of the multiple blocks including multiple characters, and determine an inclination of each of the multiple characters included in each of the multiple blocks, a block processing unit configured to detect a change point block, the change point block being a block including characters having an inclination included in a first inclination interval, a number of the characters being equal to or larger than a first threshold, and including characters having an inclination included in a second inclination interval, a number of the characters being equal to or larger than the first threshold, the second inclination interval being different from the first inclination interval, and a fold determining unit configured to determine that the document is folded if the change point block is detected.