Frontal camera evaluates parked vehicle orientation to determine parking space type, enabling reliable maneuver selection when ground markings are absent.
A vehicle controller predicts other vehicle trajectories using time series camera images and signal light states to generate safer navigation paths.
ML emulator replaces compute-intensive CFD models to enable accurate, energy-efficient wildfire spread prediction on resource-constrained edge devices.
A video feature extractor trained via self-supervised joint prediction of preceding and succeeding subsequences to capture temporal dynamics.
A 3D geospatial mapping method preprocesses satellite imagery to remove atmospheric clouds and shadows before generating point clouds.
A deep neural network predicts obstacle distances using LIDAR and RADAR ground truth data.
Processing circuitry extends ultrasonic wave receiving periods based on in-vehicle camera images to detect specific objects.
A neural network identifies hard example sensor data inputs by analyzing prediction inconsistencies across multiple outputs for the same scene.
Segmenting high variance regions into low variance sub-regions improves detection accuracy while reducing computational resource demands.
A parking assist apparatus calculates target vehicle speed and movement position to guide the car smoothly.
Static environment image extraction feeds trajectory planning models for autonomous vehicles.
A camera system switches between search and capture modes to maintain subject tracking.
Fusing 2D camera proposals with 3D LIDAR data resolves depth accuracy gaps while reducing sensor system complexity.
Mobile edge computing processes multi-spectral surveillance data to deliver low-latency augmented reality views despite compromised connectivity.
Deep convolutional neural networks detect Sargassum in multispectral satellite imagery, resolving nearshore detection accuracy issues caused by sensor noise.
Applies location-based occlusion penalties to visibility flags, resolving false negatives from obscured pedestrians in ADAS.
A detection device calculates pedestrian distance and sets a projected existence range to alert drivers.
A traffic sign recognizing apparatus combines speed limit information with conditional data using a camera sensor and control device.
Automated machine learning module splits datasets into balanced subsets by minimizing feature distribution dissimilarity.
A vehicle control system adjusts preventive safety intervention based on automated driving information.
A method constructs an endmember spectrum database by extracting unique spectra from mixed pixel data and registering them based on similarity thresholds.
Calculating synthetic transmittance images corrects satellite spectral reflectance for methane absorption, resolving accuracy limits in plume detection.
A processor calculates class-specific likelihood thresholds using stored tolerance information to optimize vehicle object detection.
A localization system classifies environmental features into semantic categories to compare only matching map data elements.
A computer method classifies objects using neural networks and combines this data with sensor readings to determine distance.
Binarizes HOG descriptors to cut data volume and enable real-time collision detection.
An object identification device generates score vectors from image features and classifies them into clusters using a dedicated determiner.
A multi-sensor apparatus determines road sign visibility states using vehicle sensor data.
A camera system adjusts exposure values to capture dark and normal images for object identification.
A vehicle driving assist device calculates risk degrees for oncoming moving bodies to perform preliminary collision avoidance control.
Map information identifies blind areas to prevent erroneous congestion relief determination when leading vehicles are undetectable by sensors.
A machine learning system conditions trajectory predictions on behavior signals to improve accuracy.
A hitch assist system autonomously aligns a vehicle hitch ball with a trailer coupler using sensor feedback.