Scene Recognition TRIZ Cases

A vanishing-point detection area adjusts exposure to separate distant headlights and taillights without overexposure in dark driving.

Mapped static objects let autonomous vehicles estimate real-time sensor perception range and adjust driving under changing weather conditions.

Multiple feature-point patterns matched to outside illuminance help automatic parking recognize a designated space despite lighting changes.

When a lead vehicle blocks the camera view, map data and other non-imaging cues keep road type detection available for expressway platooning.

A stored base graph for static elements and selective modeling of key dynamic agents cuts compute load while improving vehicle behavior prediction.

Image-based obstacle size and category recognition guides differentiated parking maneuvers to improve accuracy and protect the vehicle.

Selective annotation of uncertain segmentation pixels improves rare feature classification while reducing supervision in autonomous vehicle training.

By matching detected signs to the vehicle's lane position, this case improves speed limit recognition where different lanes show different limits.

A shared-backbone multi-task model and sensor fusion improve vehicle subtype and attribute recognition while limiting AV processing complexity.

Predicted-route risk zones focus object detection on curved-road hazards, improving driving control response without wasting computing power.

Dynamic deceleration thresholds use object and collision probabilities to cut false pedestrian and cyclist impact alarms on rough roads.

CAN driving data and location-aware thresholds improve impaired driving detection reliability and enable vehicle intervention when deviation rises.

Fusing learned-model outputs with algorithmic sensor processing improves object location and attribute detection while staying reliable in edge cases.

Overlapping camera and LiDAR views reduce close-range blind spots and occlusion, improving object classification for autonomous vehicle maneuvers.

Lane-edge modeling and vehicle trajectory tracking help driver assistance adapt when nearby vehicles occupy one or multiple lanes.

An optical system shifts high-resolution regions to key rear and side views, cutting camera count while preserving wide vehicle imaging coverage.

A stacked sensor layout separates the pixel array, converter, and DSP to limit computation noise and preserve in-vehicle image quality.

Cross-checking LiDAR-only fusion tracks with camera and radar data removes false object detections and helps prevent vehicle miscontrol.

Grid mapping and clustering fuse ultrasonic TOF and camera data to cut processor load and improve parking obstacle identification.

Fused LiDAR and image features help detect road hazards and surface profiles at distance, improving autonomous vehicle navigation in varied weather.

Combining rearview, driver-monitoring, and forward-view cameras in one mirror assembly cuts installation complexity while supporting ADAS functions.

Fusing image and LiDAR features in a transformer improves 3D hazard detection at distance and in adverse weather for autonomous driving.

Grid-mapped fusion of ultrasonic TOF and camera data cuts processor load while improving parking obstacle identification accuracy.

Scene graphs over semantic keypoints link traffic signals to travel lanes without manual HD maps, improving scalable AV intersection decisions.

Lane-aware cut-in and cut-out detection reduces false lateral-movement judgments and supports safer automated vehicle control.

When position sensing fails, historical position, illuminance, and camera direction data guide autonomous restart of moving-object control.

Switching between route history and real-time environment data by travel section improves autonomous route generation in complex areas.

Fusing associated-region lane attributes with onboard sensor data improves road perception accuracy and reduces lane conflicts in autonomous driving.

Predicting where a traffic participant will stop behind an obstacle enables earlier collision prevention with fewer unnecessary activations.

Detects two-wheeled vehicles moving between the host vehicle and an obstacle, then warns the driver or applies braking before cut-in collision risk rises.

A two-color mark with an extended straight boundary improves hitch angle detection accuracy while reducing image processing load.

Visibility maps and existence probabilities help vehicle trackers handle occlusion and sparse data for more robust decisions.

Pixel and object descriptors match common objects across frames, cutting map data volume while improving autonomous vehicle localization.

Adaptive weighting of scenario-specific driving datasets reduces overfitting and stabilizes autonomous driving model performance.

Camera-based control reduces lane-departure warnings and steering intervention when drivers intentionally pass slow or turning vehicles ahead.

External gesture and RF authentication let autonomous vehicles follow authorized traffic control commands in unusual road situations.

Cylindrical projection reduces fisheye distortion in AVM camera feeds, improving DNN object recognition and gap distance calculation.

Multiple cameras estimate a preceding vehicle's angle and lateral speed, helping route control anticipate cut-ins and keep driving stable.