This invention is related to a method to improve the performance of a deep neural network (10) for the identification of a segmentation target (111) in a medical image (12, 110), comprising the steps of performing n training steps on said deep neural network (10) for the identification of said region of interest on two different representations (13, 14) of the same segmentation target (111), said representations (13,14) being definitions of the same segmentation target (111).

A circular sign candidate extraction device includes: a memory; and a processor coupled to the memory, the processor being configured to perform processing, the processing including: detecting a circle from a captured image; specifying an annular region surrounded by the detected circle and a concentric circle, the concentric circle being a circle having a radius different from the detected circle; setting one or more pixels among pixels included in the annular region as determination pixels; and extracting a circular sign candidate from the detected circle in accordance with comparison between a color of the determination pixel and a predetermined color.

A lane line recognition method, a lane line recognition device and a non-volatile storage medium. The method includes obtaining a first image which is a road image; dividing the first image from a middle of the first image to determine a first sub-image at a left part of the first image and a second sub-image at a right part of the first image; and performing respectively a recognition operation on the first sub-image and the second sub-image to determine a first lane line in the first sub-image and a second lane line in the second sub-image.

A method and apparatus for identifying an object are disclosed. The method includes: performing linear feature detection on an image to be identified by using a linear feature detecting method to obtain detected linear features, wherein the linear feature detection method transforms detection of linear features in an image space to detection of extremal points in another space and assigns larger weights to continuous image points than to discrete image points during the transformation by using a continuous cluster factor; and identifying an object to be identified from the detected linear features by considering characteristics of the object to be identified. The method and apparatus for identifying an object of the invention, when used to detect and identify weak linear objects in high resolution remote sensing images, can effectively suppress the system noise and ambient noise, thereby successfully identifying the interested object and avoiding false alarms. Moreover, short line segments can also be identified.

Various embodiments of the present disclosure can include systems, methods, and non-transitory computer readable media configured to receive an input image associated with a test object. A set of edges are identified in the input image. A set of circles are identified based on the set of edges. A subset of circles is selected from the set of circles. The subset of circles is matched to a set of reference circles associated with a reference object. An authentication score is generated for the test object based on the matching of the subset of circles to the set of reference circles.

A system for assisting in aligning a vehicle for hitching with a trailer includes a vehicle steering system, an imager mounted with and directed to a rear of the vehicle and outputting image data, and a controller. The controller receives the image data, applies a location identifier of a vehicle hitch ball to the image data, and identifies a coupler of the trailer within the image data. The controller also outputs a steering control signal to the steering system in reversing of the vehicle to align the location identifier of the vehicle hitch ball in the image data with the coupler of the trailer.

Method and system for reading and validating identity documents that involves acquiring an image of a first and/or a second side of an identity document using a camera of a portable device; recognizing whether MRZ characters exist or are readable in the acquired image; if said MRZ characters are readable or do exist reading them obtaining a pre-identified document, or otherwise, detecting a series of local points of interests and their positions on the image calculating descriptors or vectors; and identifying the type or model of said identity document, starting by correcting perspective distortions caused by a bad relative position of the identity document with respect to the camera for obtaining a corrected and substantially rectangular image of the first and/or second side of the document at a predetermined scale which is used to perform, automatically, said identification of the identity document type or model and to automatically read and identify text and non-text information included in said corrected and substantially rectangular image.

An apparatus and method to detect a line segment or arc using Hough transform. A Hough transform unit performs contour extraction on brightness image data to generate contour image data, with pixels having a pixel value of 0 to 255, performs the Hough transform on points in the contour image data, and counts additional values represented by pixel values of points in the contour image data in a Hough table. The Hough transform unit performs contour extraction on first to third component data to generate first to third contour data with pixels having a pixel value of 0 to 255, performs the Hough transform on points in the first to third contour data, and counts additional values represented by pixel values of points in the first to third contour data in the Hough table. The detection unit comprehensively evaluates the counts to detect a line segment or arc.

Systems and methods are provided for producing a road layout model. The method includes capturing digital images having a perspective view, converting each of the digital images into top-down images, and conveying a top-down image of time t to a neural network that performs a feature transform to form a feature map of time t. The method also includes transferring the feature map of the top-down image of time t to a feature transform module to warp the feature map to a time t+1, and conveying a top-down image of time t+1 to form a feature map of time t+1. The method also includes combining the warped feature map of time t with the feature map of time t+1 to form a combined feature map, transferring the combined feature map to a long short-term memory (LSTM) module to generate the road layout model, and displaying the road layout model.

A method for barcode detection, barcode detection system and program. In method for detecting a barcode, an information processing apparatus executes the steps of: detecting a plurality of blobs from an image; for each detected blob, obtaining imaginary lines bisecting the detected blob perpendicular at the middle of the detected blob, determining at least one cluster of the imaginary lines in a feature space according to slopes and positions of the imaginary lines and grouping the blobs in one cluster to a rectangular shape corresponding to the barcode.