Embodiments of the present application are an image fusion method and a bifocal camera. The method includes: acquiring a thermal image captured by the thermal imaging lens and a visible light image captured by the visible light lens; determining a first focal length when the thermal imaging lens captures the thermal image and a second focal length when the visible light lens captures the visible light image; determining a size calibration parameter and a position calibration parameter of the thermal image according to the first focal length and the second focal length; adjusting a size of the thermal image according to the size calibration parameter, and moving the adjusted thermal image to the visible light image according to the position calibration parameter for registration with the visible light image, to obtain to-be-fused images; and fusing the to-be-fused images to generate a bifocal fused image.

A deep learning machine includes a classification unit having a labeling criterion and configured to label input data according to the labeling criterion, a conversion unit configured to integerize input data labeled as a first type requiring integerization among the input data labeled by the classification unit, a first learning data unit configured to receive the input data of the first type integerized through the conversion unit and to infer output data, and a second learning data unit configured to receive input data labeled as a second type requiring no integerization and to infer the output data.

The invention relates to a traffic flow machine-learning modeling system and method applied to vehicles. The system includes: a target fusion module configured to perform target fusion on radar measurement data and camera measurement data, and output target attribute information; a lane line model module configured to output an original lane line model based on the camera measurement data; a target selection module configured to determine a position of a lane where each target is located based on the target attribute information output by the target fusion module and the original lane line model output by the lane line model module, and output the target; and a traffic flow calculation module configured to model the vehicle position by using a clustering algorithm based on the output of the target fusion module, the output of the target selection module, and the output of the lane line model module, and output a traffic flow-based lane line model. According to the invention, accurate lane line parameters can be obtained, and related lane flow characteristic information can be provided.

On-floor obstacle detection using an RGB-D camera is disclosed. An obstacle on a floor is detected by receiving an image including depth channel data and RGB channel data through the RGB-D camera, estimating a ground plane corresponding to the floor based on the depth channel data, obtaining a foreground of the image corresponding to the ground plane based on the depth channel data, performing a distribution modeling on the foreground of the image based on the RGB channel data to obtain a 2D location of the obstacle, and transforming the 2D location of the obstacle into a 3D location of the obstacle based on the depth channel data.

A system for selectively amplifying audio signals may include a microphone configured to capture sounds from an environment of a user. The system may also include a processor programmed to: receive audio signals representative of the sounds captured by the microphone; cause selective conditioning of at least one audio signal received by the microphone from a region associated with the recognized individual; and cause transmission of the at least one conditioned audio signal to a hearing interface device configured to provide sound to an ear of the user.

This application relates to a face image processing method, apparatus, computer device, and storage medium. The method includes acquiring a first face image and a second face image, the first face image and the second face image being images of real faces; generating a first updated face image with non-real face image characteristics based on the first face image; adjusting color distribution of the first updated face image according to color distribution of the second face image to obtain a first adjusted face image; acquiring a target face mask of the first face image, the target face mask being generated by randomly deforming a face region of the first face image; and blending the first adjusted face image and the second face image according to the target face mask to obtain a target face image. Accordingly, a diversity of target face images can be generated.

Systems and methods are described for mobile and augmented reality-based depth and thermal fusion scan imaging. Some embodiments of the present technology use sophisticated techniques to fuse information from both thermal and depth imaging channels together to achieve synergistic effects for object recognition and personal identification. Hence, the techniques used in various embodiments provide a much better solution for, say, first responders, disaster relief agents, search and rescue, and law enforcement officials to gather more detailed forensic data. Some embodiments provide a series of unique features including small size, wearable devices, and ability to feed fused depth and thermal streams into AR glasses. In addition, some embodiments use a two-layer architecture for performing device local fusion and cloud-based platform for integration of data from multiple devices and cross-scene analysis and reconstruction.

Systems and methods are provided for multi-modal test-time adaptation. The method includes inputting a digital image into a pre-trained Camera Intra-modal Pseudo-label Generator, and inputting a point cloud set into a pre-trained Lidar Intra-modal Pseudo-label Generator. The method further includes applying a fast 2-dimension (2D) model, and a slow 2D model, to the inputted digital image to apply pseudo-labels, and applying a fast 3-dimension (3D) model, and a slow 3D model, to the inputted point cloud set to apply pseudo-labels. The method further includes fusing pseudo-label predictions from the fast models and the slow models through an Inter-modal Pseudo-label Refinement module to obtain robust pseudo labels, and measuring a prediction consistency for the pseudo-labels. The method further includes selecting confident pseudo-labels from the robust pseudo labels and measured prediction consistencies to form a final cross-modal pseudo-label set as a self-training signal, and updating batch parameters utilizing the self-training signal.

Embodiments automatically assess, e.g., quantify dimensions of, cracks in real-world objects. Amongst other examples, such functionality can be used to identify structural problems in bridges and buildings. An example implementation maps pixels in an image of a real-world object to corresponding points in point cloud data of the real-world object. In turn, a patch in the image data that includes a crack is identified by processing, using a classifier, the pixels with the corresponding points mapped. Pixels in the patch that correspond to the crack are then identified based on one or more features of the image. Real-world dimensions of the crack are determined using the identified pixels in the patch corresponding to the crack.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for the generation and use of a surfel map with semantic labels. One of the methods includes receiving a surfel map that includes a plurality of surfels, wherein each surfel has associated data that includes one or more semantic labels; obtaining sensor data for one or more locations in the environment, the sensor data having been captured by one or more sensors of a first vehicle; determining one or more surfels corresponding to the one or more locations of the obtained sensor data; identifying one or more semantic labels for the one or more surfels corresponding to the one or more locations of the obtained sensor data; and performing, for each surfel corresponding to the one or more locations of the obtained sensor data, a label-specific detection process for the surfel.