Scene Recognition TRIZ Cases

Parallel pipelines precompute static road attributes and combine them with dynamic edge data to speed self-driving car path graph generation.

Human speech, gaze, and gestures label relevant external objects in vehicle sensor data, improving ANN training and driving decisions.

Separate prediction and analytical models are combined to help robotic vehicles avoid dynamic and static obstacles in crowded spaces.

Natural language queries and iterative labeling help vehicle vision models find rare driving scenarios and improve prediction accuracy.

A generic detector filters sign candidates from camera and lidar data, then selective parallel classifiers improve road sign recognition for autonomous driving.

A trained neural network estimates mobile platform controller variables from 2D object features, avoiding noisy 3D projection and manual tuning.

Ground-projected driving paths help following vehicles and pedestrians anticipate self-reversing maneuvers using camera-based trajectory updates.

A single processing unit combines environment estimation, route generation, and target motion control to cut delays in autonomous driving.

Alternating vehicle radiation sources during camera capture cuts shadows and improves image contrast for more reliable object detection.

Driver attention data adjusts automated longitudinal control when traffic signals cannot be considered, improving urban driving comfort and safety.

Separate prediction and analytical models score dynamic and static obstacles, improving robotic vehicle collision avoidance in crowded spaces.

ML-based confidence scoring aggregates clustered detections into bounded probabilities, reducing false or missed objects in autonomous driving.

Aggregated coverage values and soft labels turn clustered detections into probability-based confidence scores, reducing false or missed objects.

When pathway or surrounding images change, detected change information triggers model retraining to preserve moving-object detection accuracy.

Camera visual cues fused with LIDAR help autonomous vehicles predict bicycle turns earlier and react more safely.

Error and uncertainty models turn sensor and vehicle data into collision probabilities, improving autonomous driving safety decisions.

Map data is segmented and adjusted by vehicle capability and surroundings to improve navigation accuracy without slowing real-time updates.

By turning high-dimensional sensor data into scene-element tokens, the model preserves context and improves multi-agent trajectory prediction.

Doppler velocity data from coherent LiDAR helps classify pedestrians and vehicles by motion pattern, improving AV path decisions.

Multiple traffic light images are screened and fused with time-series classification to improve autonomous driving recognition in weather and backlighting.

Taillight recognition triggers autonomous stopped-vehicle avoidance and path planning, reducing manual takeover and lane-blocking delays.

LIDAR detects pillars and vehicles while a camera reads slot markings to map indoor parking areas without markers or communication modules.

Dual upper and lower cameras compare bird's-eye road views to detect hidden rear obstacles and warn of truck collision risk.

A stereoscopic object projected along the travel lane makes lane departure risk more intuitive than planar boundary images.

Weighted occupant confirmation helps activate driving assistance when sensor data is uncertain, improving function availability and safety.

Predictive optimal control uses camera and LiDAR data to keep automated vehicles on planned trajectories despite delays and model mismatch.

UKF-based fusion of 2D camera detections and vehicle motion predicts 3D positions of static traffic objects with faster convergence and lower compute.

An edge server handles vehicle MR object detection near the camera to cut latency and bandwidth while keeping reality devices lightweight.

Targeted marker illumination adjusts light direction, range, and intensity so onboard cameras can recognize markers reliably in dark zones.

A split-screen digital rear mirror combines surround views and object alerts to shorten gaze shifts when stopping or parking.

Fusing satellite road features with probe-based lane groups helps classify unmarked or changing road sections for accurate autonomous vehicle trajectories.

Rear-image analysis detects adjacent vehicles and estimates distance and speed to guide safer lane changes with less driver guesswork.

Filters and scores occluded regions so vehicle trajectory planning can focus on likely hazards without processing every hidden area.

OCR and an LLM turn detected traffic sign text into timely spoken guidance, reducing driver cognitive load and distraction.

Obstacle uncertainty is converted into a confidence-based shape so the vehicle can plan safer trajectories without evaluating many separate predictions.

Angled image scanning finds an object's ground contact point with lower processing load, improving real-time vehicle control accuracy.

Registers the vehicle stop position during automatic parking to correct incline-related image shift and avoid obstacle overlap.

Onboard audio, sensor, and vehicle data are fused to auto-annotate environmental data, cutting manual labeling time for object recognition training.

Uses prior proceed-stop decisions with speed and position data to keep autonomous traffic light control stable when signals are briefly obscured.

When forward view falls short at speed, the vehicle increases dropback distance to restore sensor range and maintain autonomous driving.

By comparing occupant gaze with external vehicle perception, this case cuts unnecessary warnings and improves attentiveness assessment.

Moving visual links between person nodes make gratitude chains visible, reducing display overload while encouraging reciprocal kindness.

By shifting distortion-corrected image regions away from guardrails, this case improves vehicle target recognition coverage and accuracy.

Tracks pedestrian crossing intent, non-verbal cues, and audible cues to detect adversarial behavior toward autonomous vehicles.

Uses unified coordinates and multi-target relationships to estimate sensor axis deviation and enable automatic correction in vehicle monitoring.

By correcting exit-side lane data with the opposite lane, this case stabilizes lane-keeping on last-lane branch roads.

When many obstacles crowd the scene, clustered overlay marks reduce display clutter while preserving driver caution cues inside the cabin.

Mobile video screens on autonomous vehicles update ads in real time to target nearby audiences while reducing fixed-media waste and operator cost.

Rear mirror alerts linked to a backup camera help drivers detect objects behind the vehicle and confirm hazards on the in-cabin video display.

Multi-sensor detection verifies a registered user’s approach, gesture, or voice to open a vehicle trunk hands-free from a safe distance.