Scene Recognition TRIZ Cases

Sensor-verified alternate edges correct odometry drift and non-traversable links in waypoint maps, improving robot localization and path planning.

Sensors detect load protrusions so the footprint only expands where needed, improving route judgment while avoiding collisions and detours.

Camera-based state recognition lets mixed warehouse devices coordinate through control signals without interface changes, cutting integration time and downtime.

Separating in-field from out-of-field obstacle detections helps autonomous work vehicles avoid unnecessary maneuvers and preserve work efficiency.

A UAS combines methane concentration and weather data along downwind flight paths to detect leaks and estimate emission rates across gas infrastructure.

Aerial image analysis and onboard sensor feedback let an autonomous tractor reroute around changing field obstacles with less onboard complexity.

Multi-modal sensor data across timesteps removes dynamic objects and reconstructs detailed 3D scenes for more realistic robotic simulation.

Multi-modal sensor data is used to remove dynamic objects and rebuild occluded 3D regions with coarse-to-fine detail for robotic simulation.

Image recognition drives API commands that coordinate AMRs and other conveyance devices without changing existing warehouse control systems.

Uses escape history, obstacle height, and selective neural classification to identify doorsills and reduce robot sweeping delays.

Dual delay compensation combines homography image correction and MPC control prediction to reduce distortion and stabilize remote vehicle operation.

A three-phase RNN training scheme uses exponential decay to handle sparse video labels, preserve state relevance, and cut memory use.

Location-based obstacle handling separates targets inside and outside the field, avoiding unnecessary collision maneuvers while preserving work efficiency.

A UAV combines downwind trace-gas sensing with meteorological data to localize methane leaks and estimate emission rates more efficiently.

Blockchain identity, token authorization, and distress replacement help multi-drone missions stay secure and reliable in adversarial conditions.

Confidence-ranked object candidates cut dialogue turns by confirming the intended target from camera images and user directives.

By processing image regions based on UAV movement direction, this case speeds specific object detection and improves flight safety.

Image-based distance sensing lets a UAV adapt flight path and sensor settings to keep performers or objects in one field of view.

Virtual influence zones around moving objects predict path intersections and warn of deck collision or environmental exposure risks.

Sequential radio wave irradiation and reception timing data identify vehicles or other objects without onboard operation devices.

Predicting operator intent from input and sensor data helps remote vehicles ride through jitter and dropouts without false lost-link triggers.

Orientation vectors and onboard soil images reveal implement wear and state, allowing farming vehicles to adjust modes before damage occurs.

Determines stop positions that keep an autonomous mobile object within a pedestrian's visual field, improving safety in shared spaces.

A mobile object selects a stop position within an approaching traffic participant's visual field to improve recognition and stopping safety.

Partitioning a machine learning model into subgraphs across process domains reduces memory thrashing and speeds inference.

Rich sensor inputs are fused into blackboard images so autonomous machinery can map evolving context to real-time control actions.

Combining depth and height maps helps robots filter false obstacles from binocular vision and avoid collisions more accurately.

Relevance maps and a secondary classifier validate ANN image classifications, helping detect misclassifications in automotive object detection.

Traffic light positions and road-center detection localize intersection centers without HD maps, improving lane-type recognition for autonomous driving.

Centralized drone management uses role-based authorization, real-time tracking, and distress handling to coordinate multiple drones securely.

Pre-turn oscillation signals an autonomous robot's intended avoidance path, helping nearby people recognize direction changes earlier.

Height-direction point distribution helps distinguish light source noise from real obstacles, improving mobile object route determination.

Height-based point cloud analysis separates light source noise from obstacle candidates, improving mobile object route selection.

Reduced-overlap UAV image capture and selective frame processing cut mapping load while enabling near real-time orthomosaic feedback.

Ranks candidate stop areas by traffic flow, pedestrian activity, distance, and size to reduce interference during autonomous robot stops.

Sensor-driven influence models predict path intersections and environmental hazard zones before moving objects collide.

Sensor and camera models detect implement wear and state from pull vectors and soil images, enabling automatic vehicle mode changes.

Multi-sensor temporal models improve pedestrian gesture and attribute detection, helping vehicles respond more safely in high-risk scenes.

Groups vehicles by map position and travel direction to prioritize simultaneous remote support requests and reduce operator burden.

Classifying detected objects by priority improves position-related information and raises autonomous movement control accuracy.

High-speed crop-row imaging with AI and onboard lighting detects pests and diseases accurately for timely, targeted treatment.

Onboard cameras, AI, and vehicle lighting maintain pest and disease detection accuracy at high field inspection speeds, day or night.

Multi-sensor fusion and real-time learning help a two-wheeled robot adapt to dynamic environments while supporting SLAM and obstacle avoidance.

Temporary license control links multiple users to a shared autonomous device for remote shooting, access management, and content delivery.

A monitoring unit learns occupant behavior and sends drones on predicted paths to inspect property areas before alarm events occur.

Obstacle classification from onboard and facility sensors lets autonomous movers adjust defense space before starting, reducing collision risk.

LiDAR flags obstacle candidates, then the camera images only those targets to cut processing load while keeping farm machine navigation accurate.

Radar, camera, and 3D context help agricultural vehicles ignore crops, reflections, and terrain changes without missing real obstacles.

A projected vertical light line reveals floating obstacles by segment changes, helping self-guiding machines judge distance and avoid collisions.

Dynamic detection regions matched to robot mode, region, and speed improve warehousing robot braking accuracy under ambient light interference.