Abstract: Provided is a method and system of tracking partially occluded objects using an elastic deformation model. According to an exemplary method and system, partially occluded vehicles are detected and tracked in a scene including side-by-side drive-thru lanes.

Abstract: Methods, systems, and processor-readable media for the detection and classification of license plates. In an example embodiment, an image of a vehicle can be captured with an image-capturing unit. A license plate region can then be located in the captured image of the vehicle by extracting a set of candidate regions from the image utilizing a weak classifier. A set of candidate regions can be ranked utilizing a secondary strong classifier. The captured image can then be classified according to a confidence driven classification based on classification criteria determined by the weak classifier and the secondary strong classifier.

Abstract: A detection-based segmentation-free method and system for license plate recognition. An image of a vehicle is initially captured utilizing an image-capturing unit. A license plate region is located in the image of the vehicle. A set of characters can then be detected in the license plate region and a geometry correction performed based on a location of the set of characters detected in the license plate region. An operation for sweeping an OCR across the license plate region can be performed to infer characters with respect to the set of characters and locations of the characters utilizing a hidden Markov model and leveraging anchored digit/character locations.

Abstract: Methods, systems, and processor-readable media for detecting the side window of a vehicle. A spatial probability map can be calculated, which includes data indicative of likely side window locations of a vehicle in an image. A side window detector can be run with respect to the image of the vehicle to determine detection scores. The detection scores can be weighted based on the spatial probability map. A detected region of interest can be extracted from the image as extracted image patch. An image classification can then be performed with respect to the extracted patch to provide a classification that indicates whether or not a passenger is in the vehicle or no-passenger is in the vehicle.

Abstract: A video sequence can be continuously acquired at a predetermined frame rate and resolution by an image capturing unit installed at a location. A video frame can be extracted from the video sequence when a vehicle is detected at an optimal position for license plate recognition by detecting a blob corresponding to the vehicle and a virtual line on an image plane. The video frame can be pruned to eliminate a false positive and multiple frames with respect to a similar vehicle before transmitting the frame via a network. A license plate detection/localization can be performed on the extracted video frame to identify a sub-region with respect to the video frame that are most likely to contain a license plate. A license plate recognition operation can be performed and an overall confidence assigned to the license plate recognition result.

Abstract: A method for detecting a vehicle running a stop signal includes acquiring at least two evidentiary images of a candidate violating vehicle captured from at least one camera monitoring an intersection. The method includes extracting feature points in each of the at least two evidentiary images. The method includes computing feature descriptors for each of the extracted feature points. The method includes determining a correspondence between feature points having matching feature descriptors at different locations in the at least two evidentiary images. The method includes extracting at least one attribute for each correspondence. The method includes determining if the candidate violating vehicle is in violation of running the stop signal using the extracted attribute.

Abstract: A method for detecting a vehicle running a stop signal positioned at an intersection includes acquiring a sequence of frames from at least one video camera monitoring an intersection being signaled by the stop signal. The method includes defining a first region of interest (ROI) including a road region located before the intersection on the image plane. The method includes searching the first ROI for a candidate violating vehicle. In response to detecting the candidate violating vehicle, the method includes tracking at least one trajectory of the detected candidate violating vehicle across a number of frames. The method includes classifying the candidate violating vehicle as belonging to one of a violating vehicle and a non-violating vehicle based on the at least one trajectory.

Abstract: A method and system for domain adaptation based on multi-layer fusion in a convolutional neural network architecture for feature extraction and a two-step training and fine-tuning scheme. The architecture concatenates features extracted at different depths of the network to form a fully connected layer before the classification step. First, the network is trained with a large set of images from a source domain as a feature extractor. Second, for each new domain (including the source domain), the classification step is fine-tuned with images collected from the corresponding site. The features from different depths are concatenated with and fine-tuned with weights adjusted for a specific task. The architecture is used for classifying high occupancy vehicle images.

Abstract: Automated low-complexity video-based vehicle speed estimation is described, that operates within the video stream to screen video sequences to identify and eliminate clear non-violators and/or identify and select potential violators within a multi-layer speed enforcement system, in which deeper layers provide enhanced accuracy on selected candidate (speeding) vehicles. Video motion vector clusters corresponding to a vehicle are identified and tracked across multiple frames of captured video. Movement of the motion vector clusters translated from pixels per second to real speed (e.g. miles per hour) to determine whether the vehicle was speeding. Estimated speed data is added to the video stream data is metadata, and video segments of candidate speeding vehicles are stored and/or transmitted for subsequent review (e.g. automated or manual).

Abstract: This disclosure provides a method and system for automated sequencing of vehicles in side-by-side drive-thru configurations via appearance-based classification. According to an exemplary embodiment, an automated sequencing method includes computer-implemented method of automated sequencing of vehicles in a side-by-side drive-thru, the method comprising: a) an image capturing device capturing video of a merge-point area associated with multiple lanes of traffic merging; b) detecting in the video a vehicle as it traverses the merge-point area; c) classifying the detected vehicle associated with traversing the merge-point area as coming from one of the merging lanes; and d) aggregating vehicle classifications performed in step c) to generate a merge sequence of detected vehicles.

Abstract: A system and method for detecting a seat belt violation in a vehicle. The method may include receiving an image including a region of interest where a seat belt is visible if the seat belt is buckled. One or more first features may be extracted from the image. The one or more first features may be compared to one or more second features extracted from a plurality of sample images. It may be determined whether the seat belt is buckled in the image based upon the comparison.

Abstract: Methods and systems for recognizing a license plate character. Synthetic license plate character images are generated for a target jurisdiction. A limited set of license plate images can be captured for a target jurisdiction utilizing an image-capturing unit. The license plate images are then segmented into license plate character images for the target jurisdiction. The license plate character images collected for the target jurisdiction can be manually labeled. A domain adaptation technique can be utilized to reduce the divergence between synthetically generated and manually labeled target jurisdiction image sets. Additionally, OCR classifiers are trained utilizing the images after the domain adaptation method has been applied. One or more input license plate character images can then be received from the target jurisdiction. Finally, the trained OCR classifier can be employed to determine the most likely labeling for the character image and a confidence associated with the label.

Abstract: This disclosure provides vehicle detection methods and systems including irrelevant search window elimination and/or window score degradation. According to one exemplary embodiment, provided is a method of detecting one or more parked vehicles in a video frame, wherein candidate search windows are limited to one or more predefined window shapes. According to another exemplary embodiment, the method includes degrading a classification score of a candidate search window based on aspect ratio, window overlap area and/or a global maximal classification.

Abstract: Multi-stage vehicle detection systems and methods for side-by-side drive-thru configurations. One or more video cameras for an image-capturing unit) can be employed for capturing video of a drive-thru of interest in a monitored area. A group of modules can be provided, which define multiple virtual detection loops in the video and sequentially perform classification with respect to each virtual detection loops among the multiple virtual detection loops, starting from a virtual detection loop closest to an order point, and when a vehicle having a car ID is sitting in a drive-thru queue, so as to improve vehicle detection performance in automated post-merge sequencing.

Abstract: Methods and systems for bootstrapping an OCR engine for license plate recognition. One or more OCR engines can be trained utilizing purely synthetically generated characters. A subset of classifiers, which require augmentation with real examples, along how many real examples are required for each, can be identified. The OCR engine can then be deployed to the field with constraints on automation based on this analysis to operate in a “bootstrapping” period wherein some characters are automatically recognized while others are sent for human review. The previously determined number of real examples required for augmenting the subset of classifiers can be collected. Each subset of identified classifiers can then be retrained as the number of real examples required becomes available.

Abstract: This disclosure provides a method and system for automated sequencing of vehicles in side-by-side drive-thru configurations via appearance-based classification. According to an exemplary embodiment, an automated sequencing method includes computer-implemented method of automated sequencing of vehicles in a side-by-side drive-thru, the method comprising: a) an image capturing device capturing video of a merge-point area associated with multiple lanes of traffic merging; b) detecting in the video a vehicle as it traverses the merge-point area; c) classifying the detected vehicle associated with traversing the merge-point area as coming from one of the merging lanes; and d) aggregating vehicle classifications performed in step c) to generate a merge sequence of detected vehicles.

Abstract: A method for training a vehicle detection system used in a street occupancy estimation of stationary vehicles. The method includes defining first and second areas on an image plane of an image capture device associated with monitoring for detection of vehicles. The method includes receiving video-data from a sequence of frames captured from the image capture device. The method includes determining candidate frames that include objects relevant to a classification task in the second area. The method includes extracting the objects from the candidate frames, extracting features of each extracted object, and assigning labels to the each extracted object. The method includes training at least one classifier using the labels and extracted features. The method includes using the at least one trained classifier to classify a stationary vehicle detected in the first area.

Abstract: Methods and systems for bootstrapping an OCR engine for license plate recognition. One or more OCR engines can be trained utilizing purely synthetically generated characters. A subset of classifiers, which require augmentation with real examples, along how many real examples are required for each, can be identified. The OCR engine can then be deployed to the field with constraints on automation based on this analysis to operate in a “bootstrapping” period wherein some characters are automatically recognized while others are sent for human review. The previously determined number of real examples required for augmenting the subset of classifiers can be collected. Each subset of identified classifiers can then be retrained as the number of real examples required becomes available.

Abstract: A method for updating an event sequence includes acquiring video data of a queue area from at least one image source; searching the frames for subjects located at least near a region of interest (ROI) of defined start points in the video data; tracking a movement of each detected subject through the queue area over a subsequent series of frames; using the tracking, determining if a location of the a tracked subject reaches a predefined merge point where multiple queues in the queue area converge into a single queue lane; in response to the tracked subject reaching the predefined merge point, computing an observed sequence of where the tracked subject places among other subjects approaching an end-event point; and, updating a sequence of end-events to match the observed sequence of subjects in the single queue lane.

Abstract: A method for determining parking availability includes receiving video data from a sequence of frames taken from an image capture device monitoring a parking area. The method includes detecting at least one object located in the parking area. The method includes determining boundaries of the parking area. The boundaries include at least an inner boundary relative to the image capture device and an outer boundary relative to the image capture device. The outer boundary is substantially parallel to the inner boundary. The method further includes computing a length of at least one of the object and a space between objects using an object pixel for each of the inner and outer boundaries. Using the computed length, The method includes determining a parking availability in the parking area. The method includes outputting a notice of the parking availability to a user.