Scene Recognition TRIZ Cases

A reference path change unit adjusts lateral trajectory based on environmental distance data.

A dynamic narrow AI agent allocation method selects context-specific subsets of agents to process vehicle sensor data.

Generative adversarial networks generate target shadow backgrounds to resolve accuracy complexity trade-offs in synthetic aperture radar identification.

A control unit tracks a vertical trailer edge position from consecutive camera images to update the digital mirror display section.

Edge-mounted vehicle cameras capture image data for real-time double parking detection using semi-supervised learning models.

A vehicle object tracking system selects algorithms based on environmental settings and available computational resources.

A pedestrian movement prediction method uses segmented velocity components to determine direction for autonomous vehicles.

Binary mask segmentation and temporal smoothening filter noise to improve traffic sign detection accuracy under varying visual conditions.

Machine learning models process satellite images to classify land habitats, reducing manual survey resource consumption.

A self-adaptive image processing method converts RGB frames to HSV color space and segments blocks using hue and saturation statistics.

A neural network filters sensor data to select relevant objects for vehicle-to-everything communication.

Automated image processing replaces manual surveys to generate timely diagrams, while drone verification ensures measurement precision.

A vehicle system uses an intersection reflector and image sensor to detect approaching vehicles.

Detects materials in compressed hyperspectral pixels using basis vector projection to compute detection scores directly from spectral coefficients.

A predictive turning assistant uses AI to estimate object movement patterns during vehicle turns.

Adaptive predictive modeling analyzes temporal spectral radiance changes to detect small fires and atmospheric phenomena from satellite imagery.

Knowledge graph augmentation bridges the gap between deep neural network outputs and scene context, resolving explainability issues in automated identification.

Counting frames from detection to a target position replaces complex coordinate calculations, eliminating processing delays in alarm timing.

A distance information calculation unit divides captured images into pixel blocks to compute individual distances for object detection.

A multi-camera vision system presents selected feeds in a main area and non-selected feeds in gallery thumbnails on a work vehicle display.

A deep neural network processes LIDAR point clouds to generate 3D object representations and determine rotational orientation.

A generative model learns distributions between high-resolution and low-resolution satellite images to produce synthetic time series data.

Segmenting detection tasks across sensors eliminates blind spots while maintaining manageable system complexity.

An infrared-emitting street light provides illumination that enables a vehicle computer to identify objects in low light conditions.

A vector extraction engine processes raster imagery to generate precise two-dimensional and three-dimensional linear features using digital surface models.

Camera-guided lidar processing reduces computational load by correlating classified objects with point clouds for efficient 3D modeling.

Dynamic damping reduces signal volatility in indoor radar sensors while object lifespan filtering distinguishes real people from ghost detections.

A control apparatus detects unstable loads on nearby vehicles to adjust trajectory and speed for collision prevention.

A digital scene generator creates photo-realistic 3D human models and assigns them to coordinate locations within reconstructed environments.

Image processing unit detects obstacles by comparing overlapped regions in bird's-eye-view images.

Control device calculates corrected parking position via second trajectory to prevent door obstruction and ensure alignment.

A road sign detection device uses dictionary information and Gray Level Co-Occurrence Matrix analysis to identify signs in vehicle images.

Dynamic parameter adjustment resolves fixed-radius limitations, enabling accurate object detection across varying sensor distances and point densities.

Machine learning algorithm generates mobility and velocity masks from fused LIDAR tensors to identify moving objects in real time.