Scene Recognition TRIZ Cases

Independent validation modules compare autonomous vehicle perception outputs and use consensus rules to flag errors for safer navigation.

Independent perception modules compare sensor outputs and use voting to flag discrepancies and determine more reliable ground truth.

Real-time perception checks traffic, obstacles, and ground hazards against stopping rules so an AV can confirm or change its pullover location.

Dynamic transmission patterns and array antenna selection help vehicle radar detect near and far targets while reducing unnecessary energy use.

Occupancy width on a curve guides obstacle-aware path control, helping autonomous vehicles avoid bend hazards more safely.

Real-time AR overlays combine sensor and GPS road sensing to cut display latency and improve driver awareness on snowy or muddy roads.

Anticipating adjacent vehicle cut-ins, this case uses corrected longitudinal distance to smooth speed control and maintain safe spacing.

Grid-based point cloud processing estimates free space by weighting sensor noise, availability, obstacle height, and distance for safer AV navigation.

Dynamic wait-time logic helps autonomous vehicles avoid indefinite stops at multi-way intersections while yielding safely to other road users.

Visual alerts in rearview mirrors prompt drivers to check rear camera views, improving detection of objects behind the vehicle while reversing.

Risk-factor-first alerts cut driver overload when deceleration assist and risk avoidance operate at the same time.

Driver eye tracking and disturbance detection keep head-up display object cues aligned during vehicle vibration and sudden motion.

By limiting stop line search depth to less than twice the crosswalk depth, this case cuts false detections from nearby road markings.

Uncertainty indicators guide selective reference point transformations, improving vehicle object association and fusion reliability with limited computing power.

Contrastive learning reshapes object embeddings so detectors separate same and different objects, reducing false positives without added inference cost.

Sensor-based prediction lets vehicle headlights switch before upcoming exterior lighting changes, reducing reaction delay and improving driving safety.

When lane lines or a lead vehicle disappear, control switches among lane, vehicle, and virtual lane following to keep travel stable.

Multiple QP paths with different obstacle buffers run in parallel, enabling safe autonomous driving while keeping closer to the road center.

Real-time side-mirror camera images are shown when mirrors are obstructed or turn signals activate, helping drivers see during lane changes.

Horizontal optical flow transitions let a mono camera detect drivable space and estimate distance without stationary objects or flat-road assumptions.

Virtual vehicles are placed behind detected target-lane traffic to guide smoother lane changes when following vehicles are not yet recognized.

A server wakes a parked vehicle, switches it to a movable state, and guides obstacle-aware repositioning when the driver is absent.

Opposite-lit indicator units let a following vehicle read a leader's light state accurately under sunlight for stable travel control.

LiDAR first flags obstacle candidates, then triggers camera imaging to keep autonomous farm machine detection accurate with lower processing load.

Simulated sensor data from a 3D-scanned environment trains robotic carts to recognize objects despite occlusion and changing conditions.

Multiple stereo cameras and per-camera neural networks improve small-object anomaly detection under occlusion and changing field lighting.

Multi-pass camera, LiDAR, and radar validation cuts false anomaly calls in fields, improving vehicle safety and control precision.

Time-synced camera, LiDAR, and radar data validate field anomalies across passes, improving detection robustness for agricultural vehicles.

Dual stereo cameras and separate neural networks improve detection of small, thin, and irregular field obstacles for safer vehicle control.

Dynamic ego and object motion models define perception zones that flag safety-critical objects more accurately than static detection metrics.

A movable sensing platform detects and indexes dynamic field obstacles in real time, enabling autonomous farm vehicles to navigate safely.

A unified CNN trained on labeled field images detects diverse obstructions and updates farming actions in real time to avoid damage.

Aerial image routing lets autonomous tractors detect field obstacles and update paths from onboard sensor data with less hardware.

A three-phase training flow uses exponential decay state weighting to keep video object detection accurate with sparse labels and lower training cost.

Sensor-verified alternate edges and nonlinear optimization correct odometry drift in robot waypoint maps, enabling shorter paths and better localization.

Image sensors and ML detect hidden roadside boxes, rocks, and debris so mowing vehicles can warn operators, avoid damage, and map hazards.

Image sensors and machine learning identify hidden roadside boxes and debris, helping mowing vehicles avoid damage, repairs, and safety risks.

3D object tracking and a scene encoder predict probable paths in warehouse traffic, helping autonomous forklifts avoid conflicts without excess stops.

Radio-wave target tracking combined with acoustic obstacle sensing guides autonomous movement in blocked paths without costly cameras, LiDAR, or radar.

Thermal imaging and 3D mapping help agricultural vehicles detect animals and obstacles, geofence them, and reroute safely during field work.

Voxel-grid filtering cuts 3D LiDAR point clouds by marking low-density cells unoccupied, easing SLAM and obstacle detection on material handling vehicles.

Thermal imaging and 3D mapping detect animals and obstacles, then geofence them so agricultural vehicles can avoid harm and equipment damage.

Selective depth extraction from stereo images cuts processing load while preserving accurate vehicle pose estimation for obstacle avoidance.

Voxel occupancy filtering cuts 3D LiDAR point clouds by 80-95%, easing SLAM and obstacle detection on material handling vehicles.

Visual pattern and distance detection let the UAV adjust flight and sensor settings to keep performer and performee in one field of view.

Camera-based obstacle detection lets a work vehicle stop or continue automatically, balancing safety with field work efficiency.

Image-trained obstruction detection lets farming machines identify diverse field hazards and modify treatment paths in real time to avoid damage.

Calculates operator object-detection probability from image quality and distance to flag hazard visibility loss in vehicle monitoring.

Obstacle type discrimination lets a work vehicle shorten avoidance distance for animals like birds, reducing waste travel while maintaining safety.

NeRF models compress UAV aerial imagery into accurate terrain views, cutting transmission load while supporting safer delivery navigation and simulation.