Image Enhancement TRIZ Cases

Combining PET metabolic data with CT or MRI radiomics improves metastasis prediction accuracy while reducing reliance on manual assessment.

Sensor data and ML assignment functions derive current lane descriptions from markings and nearby objects for more accurate autonomous driving.

Autoencoder-based feature compression and soft-label clustering improve wafer defect classification accuracy while reducing high-dimensional processing load.

Through-beam sensors and X-axis translation with Z-axis rotation keep vehicles at a consistent pose for more accurate vision calibration.

Diffuse illumination inside a black box images mirror-like cleaved wafers without light-source reflections, improving in-line defect detection.

Onboard sensors build a baseline cargo profile, detect securement issues and cargo shifts, and trigger corrective driving or rerouting.

A dual-group optical module combines structured light projection with diffuse illumination to cut module count, power use, and 3D sensing complexity.

A trailer camera compares a selected load point to a fixed reference and alerts the driver when cargo shifts without constant live-feed viewing.

Brightness-change sensing helps detect moving objects partly hidden by obstacles, enabling collision prevention support in blind areas.

Probabilistic occlusion maps turn LiDAR, radar, or camera blind spots into navigable risk data for better obstacle avoidance in autonomous vehicles.

Scatterometry signals trained against destructive measurements predict wafer profiles, CDs, and contours faster with less coupon loss.

Surface profile maps and etch-rate modeling guide substrate placement to reduce tilt, scrapping, and edge variation in process chambers.

Combined coaxial vertical and ring illumination improves defect detection on wire-bonded semiconductor modules with uneven surfaces.

Camera-tracked head and eye movements are correlated with in-cabin mobile use events to distinguish driver distraction from passenger phone use.

Dynamic structured light patterns improve 3D measurement of translucent surfaces by reducing diffuse noise and preserving data density.

Color-coded markers overlay detected objects and predicted paths on vehicle video to improve driver awareness of nearby collision risks.

Camera and infrared sensing detect parked car door opening cues early, letting the host vehicle adjust its path faster and more safely.

Multi-sensor 3D space tracking combines LiDAR, radar, cameras, and server reconstruction to extend sensing range and support stable vehicle control.

Camera view switches between narrow and wide angles when nearby vehicles enter blind spots, improving driver visibility and reaction time.

Computer vision locates tab root corners and reference edge lines to detect cell assembly tab misalignment with higher accuracy and speed.

Digital image correlation tracks rotor and retention ring strain without disassembly, cutting test time while preserving condition assessment accuracy.

Probability maps and trajectory lattices improve multi-agent motion prediction for autonomous navigation while limiting real-time computation.

Aerial streetlight nodes process video into object bounds and danger zones to detect near misses in real time and trigger alerts.

Thermal, NIR, and stereo imaging replace costly LIDAR to improve vehicle object detection, distance estimation, and false-positive control.

Sequential side-view camera images track object direction and speed to judge safe vehicle movement at occluded intersections beyond LiDAR range.

When GPS or site scale leaves riders misplaced, the vehicle requests user location images or coordinates and adjusts its approach for easier boarding.

Cross-sensor-trained sequential DNNs predict time-to-collision and object motion from image sequences alone, avoiding sensor inputs at deployment.

Delayed post-cleaning dirt checks and reflection-based target selection cut false LiDAR window contamination detection and liquid waste.

Frequency distribution processing of stereo range images cuts detection time while preserving road-surface and object recognition for vehicle control.

Historical pedestrian data and neural networks identify unmarked crossing hotspots so autonomous vehicles can react earlier and reduce collision risk.

Simulation-generated velocity grids train ML models to predict moving-object speeds more accurately, especially on curved trajectories.

Deep learning trained on simulated tomography data reconstructs missing wedge regions and improves atomic-level nanomaterial structure accuracy.

Weighted fusion of time-related surround views uses ORB features and vehicle motion to fill underbody blind areas in real time.

Camera image regions and occupant subtraction are combined with sun-position modeling to map cabin sunload without dedicated light sensors.

Onboard sensors detect AV issues, guide users through cleaning or minor repairs, and verify completion to keep fleet vehicles available.

Fused camera and LiDAR tracking improves surround vehicle trajectory prediction by modeling maneuvers and inter-vehicle interactions.

A protected camera and cup structure tracks low-flow bevel liquid discharge on rotating substrates to catch instability and splashing.

Lane lines extracted from on-road sensor data form a virtual target for real-time vehicle sensor calibration without rotating rigs or physical targets.

A compact lens actuator places the position sensor over base electronics to support precise focus movement with better reliability in thin devices.

Pre-calibrated fiducial markers and feature points locate objects in real-world coordinates from fisheye images with less processing.

ROI-based illuminance assessment flags low-light frames before image recognition, improving autonomous driving and parking reliability.

Hand, face, and skeletal cues enable reliable in-vehicle activity classification even when held objects are occluded or unseen.

Dual-face tab imaging with multi-frame synthesis improves layer counting and catches folded battery tabs more accurately during cell manufacturing.

Camera images classify parking lot type when GPS context is unreliable, enabling adaptive automated parking and accurate spot detection.

Image-based distance changes across two moments predict vehicle collisions without target classification, cutting computing load and storage needs.

Automated ROI tracking switches between field shifts and stage motion to correct large in-situ drift while controlling electron dose.

Depth-aware keypoints learned from monocular video improve ego-motion estimation under lighting and viewpoint changes without ground-truth labels.

Similar wafer patterns and prebuilt analysis models let inspection tools set optical conditions quickly without manual tuning or wafer images.

Image-based navigation checks next-state lateral spacing against yaw-based braking distance to keep autonomous vehicles safe and compliant.

Multiple load lock cameras detect substrate misplacement and handling defects while providing image feedback to calibrate transfer robots.