Scene Recognition TRIZ Cases

Real-time oncoming traffic, pedestrian, and signal data adjust turn deceleration timing and target speed to better match driver expectations.

Multiple classification scores combine sensor and GNSS data to flag erratic remote vehicles and improve collision avoidance responses.

Outside sensors detect obstacles at a stop position so vehicle height or stopping position can be adjusted for safe, easy passenger entry and exit.

Parallel vehicle trajectory analysis and happens-before ordering reduce false positives when identifying which vehicle initiated abnormal driving.

A motorized roof rail repositions shared ADAS modules to fit different vehicle layouts while improving sensor integration and detection accuracy.

Multiple cameras combine no-u-turn sign and right-turn lane marking detection to avoid false driver alerts when lane position is unclear.

Machine learning corrects weather- and sensor-degraded satellite images to improve remote hydrocarbon equipment monitoring and fault prediction.

Camera-selected image regions and sparse map data cut storage and processing load while preserving object-based vehicle navigation accuracy.

Vehicle odometry and obstacle sensing auto-label images to train drivable-path segmentation without manual annotation in complex urban scenes.

Bayesian maneuver classification uses vehicle position history and lane profiles to predict lane changes earlier and reduce noise-driven errors.

Forward and rear sensor fusion assesses passing safety and warns overtaking vehicles when road conditions make the maneuver unsafe.

Iterative ML and stream connection updates forecast vehicle and actor trajectories across candidate paths for more reliable AV control.

Self-attention and residual blocks encode surrounding and target vehicle data to improve driving decisions in complex traffic scenes.

Multi-region object assessment helps autonomous vehicles judge when to pass over or under foliage and small road objects without leaving the lane.

Image components are reclassified and target motion vectors are corrected from non-target movement to improve off-road tracking on uneven terrain.

Camera and radar features are fused in a BEV model to detect and classify traffic signs accurately without lidar.

Fusing cabin and exterior sensor data with neural networks improves passenger interaction detection and object awareness in autonomous vehicles.

Latency distortion in temporal sensor data is used as a signal to infer object velocity, shape, and bounding boxes for better tracking.

Obstacle kinematics weight a barrier-based control input, helping ADAS actuate braking or steering with lower computation while preventing collisions.

Measurement data is interpreted across possible vehicle sensor mountings to detect misalignment and identify the correct configuration automatically.

Chroma key extraction creates scalable blocked-camera training images with precise blockage boundaries to improve neural network detection.

A margin added to detected peripheral objects helps parking control handle narrow-space sensing errors while reducing contact risk through real-time updates.

Fused robot sensor data with redundancy information helps distinguish uncertain detections and supports safer automated control responses.

A unified radar detection head combines sensor fusion and ego-motion compensation to improve vehicle object detection accuracy with lower complexity.

Predicted offsets and yielding probability help an autonomous vehicle choose a proactive lane change path without waiting for other vehicles to yield.

Time-series lane tracking links detected signs to the correct lane, avoiding branch-road sign misclassification in driver assistance.

Trust metrics combine vehicle, driver, and situational safety data so autonomous vehicles can assess nearby traffic risk and respond safely.

Camera-based reference tracks align RADAR plots and LiDAR clusters to improve external object position and distance detection for vehicle control.

Adjusts turn deceleration at intersections using oncoming traffic, pedestrians, and signal state to better match driver expectations.

Video and telemetric sensing detect visible bump indicators early, giving drivers timely alerts and speed guidance to avoid late braking.

A stopped vehicle widens camera coverage across its front to catch pedestrians and other entering objects before following a lead car.

Dynamic display-region centering keeps both the vehicle and target parking spot fully visible during automatic parking.

When a trailer is attached, the vehicle receives its length, height, and weight to recalculate safety zones, stopping distance, and turning radius.

Self-supervised image token generation cuts annotation effort while improving environmental perception for autonomous driving control prediction.

Associating sensor detections with stored landmarks locates roadway obstacles more robustly despite calibration, localization, and map errors.

A camera overlay predicts the trailer side sweep during turns, helping drivers avoid inner-arc collisions without overly wide maneuvers.

Abnormal long-range Lidar detections are cross-checked with camera or radar data to recover road boundaries and reduce misrecognition.

Camera-based coast control reduces throttle before braking to keep safe following distance while improving ride comfort and limiting brake wear.

Radar feature grid maps and 3D occupancy mapping improve obstacle and infrastructure detection while limiting sensor system complexity.

Radar point data is turned into 3D occupancy and feature grid maps, then refined by CNN processing for more accurate vehicle environment detection.

A camera-verified boundary check helps radar-based pre-crash control suppress ghost objects caused by roadside sub-reflections.

False positive monitoring lets an autonomous vehicle raise its allowed speed by switching to a less restrictive sensor performance level.

Front-corner sensing tracks adjacent-lane objects and differential speed to warn drivers or intervene when visibility is blocked.

Detecting entry to and exit from diverging diamond interchanges lets drive assist switch control modes for accurate lane handling.

Semantic segmentation clusters point cloud regions to generate attentive anchors, improving 3D object localization and classification.

Road-surface light patterns from surrounding vehicles are recognized to predict maneuvers early and improve host vehicle hazard response.

A shared multi-task model combines detection, trajectory forecasting, and behavior intent prediction to cut processing time in autonomous driving.

By comparing camera frames captured with and without pulsed laser light, this case enables lower-cost long-range object detection.

Rear-view video analysis replaces damage-prone radar sensors to detect blind-spot vehicles and define available parking areas.