Biometric Pattern Recognition TRIZ Cases

A pet safety management system uses video analysis to detect target pets and active objects, triggering a warning device on the animal.

A computer vision system detects human actions and physical contact in combat sports using machine learning object detection algorithms.

A semantic feature extraction method generates compressed information packets by matching current frames with historical data.

Adapts a pre-trained deep neural network to specific hierarchical classes, reducing annotation time and retraining costs while maintaining high accuracy.

A human body recognition system uses three-dimensional spatial coordinates to calculate back-projection errors for iterative re-identification.

An information processing apparatus determines boundaries between partial regions in an image to estimate object movement.

COMPOSER architecture extracts keypoint data from video frames to enable group activity prediction without RGB imagery.

A person re-identification network trained on unlabeled images using block processing and random ordering to generate negative samples.

A computer system generates curated watch lists for retail stores by applying prediction models to case files and ranking users based on calculated threat scores.

A computer method generates histogram distributions to identify optimal split points for dataset division.

An image processing apparatus determines a clip region based on object positions to generate a coherent second image.

Laser scanner evaluates reflection point density distribution against anthropometric parameters to distinguish humans from background noise interference.

Unified volumetric descriptors merge separate feature sets into a single structure, resolving complexity trade-offs in object representation across variations.