Scene Recognition TRIZ Cases
- Dynamic Neural Network Layer Compression for Resource Constraints
A neural network compression system adjusts layer encoding ratios based on real-time bandwidth and thermal conditions.
- Lidar Traffic Sign Detection via Point Clustering
Clustering lidar point data into plane object clusters enables accurate traffic sign detection regardless of environmental changes.
- LIDAR Point Cloud Traffic Sign Detection for HD Map Updates
Autonomous vehicles use LIDAR point clouds to detect traffic signs and classify them via deep learning, updating HD maps without dedicated survey fleets.
- Drone Surface Inspection Using Projected Reference Image
A drone captures a projected reference image alongside an undifferentiated surface to enable precise defect location identification.
- Image Recognition Support Apparatus for Automated Label Correction
A pseudo label generation unit produces recognition data from multiple models to create new labels with high reliability.
- Environment Recognition Device for Traffic Light Identification
An environment recognition device groups target portions by luminance and relative distance to specify objects.
- Map-Assisted Object Recognition for Driver Assistance Systems
A driver assistance system uses map data to evaluate environmental sensor readings for accurate object recognition.
- Vehicular Event Recorder Buffer Segmentation
Segmented memory buffers store pre-event data, resolving conflicts between data collection completeness and limited storage capacity.
- Object Characterization via Grid Map Quantization
Quantizing high-resolution sensor samples into a weighted grid map to compute object line parameters for efficient spatial characterization.
- Travel Control Apparatus Switching to Manual Driving After Lane Line Loss
A travel control apparatus switches from lane maintenance to manual driving when road division lines become unrecognizable.
- Neural Network Captioning for Unknown Traffic Signs
A neural network generates descriptive captions for unknown traffic signs by processing image data.
- Merging Batch Normalization into Convolutional Layers
A modified convolution layer absorbs batch normalization parameters to perform inference in a single computational pass.
- Per-lane Traffic Data Collection via Video Analysis
Video capture device generates lane-specific traffic metadata to resolve per-road data inaccuracy.
- Multi-frame Moving Object Detection in Low SNR Imagery
A computing system extracts image chips along candidate motion paths to accumulate signal energy for object identification.
- Adaptive Projection Surface for Distortion-Free Surround View
A camera surround view system reprojects textures onto an adaptive projection surface calculated from sensor data.
- Hybrid Crop Monitoring System for Precision Fertilizer Application
A hybrid crop monitoring system combines satellite imagery with ground-based sensor data to generate calibrated plant metrics.
- Road Sign Content Prediction for ML Training Data
Visual attribute prediction models validate annotations via knowledge graph reasoning to reduce manual labeling time.
- Connectionist Network Image Data Processing Offsets
A connectionist network processes image data by determining individual offsets for picture elements to create a sampling grid.
- 3D Obstacle Detection Using Virtual Stixel Models
A machine-learning system generates virtual 3D stixel models from single camera images to detect obstacles and free space.
- Thermal Wake Detection via Infrared Anomaly Masking
Automated infrared imaging generates thermal anomaly masks to identify vessel wakes across vast ocean areas without manual inspection.
- Radar-Based Lane Detection Using Surrounding Vehicle Positions
A radar device detects surrounding vehicles to determine driving lane positions without relying on static road features.
- Pedestrian Detection via Object Visibility Map Fusion
An object visibility map highlights visible pedestrian features to enhance feature maps for accurate detection.
- Camera Object Validation via Homography Plane Assignment
A camera system validates detected objects by assigning image features to spatial planes using computed homographies.
- Iterative Appearance Search for Video Object Detection
Automated computer vision iteratively searches video data using reference images, reducing time required to locate missing objects.
- Offset Sensor Array for Low-Speed Impact Object Detection
Offset sensor array determines impact object shape to resolve detection accuracy limits.
- AR Head-Up Display for Contextual Driving Rule Awareness
Augmented reality head-up display filters detected traffic signs to show only applicable driving rules based on vehicle state.
- Automated Crop Phenology Image Filtering via NDVI Comparison
A computing device compares sequential NDVI values to flag and remove inconsistent crop images before model training.