Human Presence Detection via Aggregate Confidence Heatmaps

The patent merges multiple overlapping regions of interest by aggregating their confidence scores. Instead of selecting only the single highest confidence region, the system combines confidence values from all overlapping regions to create a composite confidence score, thereby improving detection accuracy while managing processing complexity through efficient aggregation algorithms.

The system implements a universal confidence aggregation mechanism that handles multiple overlapping regions simultaneously. This multi-functional approach allows the same processing framework to evaluate and combine confidence scores from any number of overlapping regions, making the system adaptable to varying detection scenarios without requiring separate processing paths.

The patent segments the confidence aggregation process into distinct stages: identifying overlapping regions, calculating individual confidence scores, aggregating scores from overlapping regions, and generating final detection results. This segmentation allows the system to manage processing complexity by breaking down the computationally intensive aggregation task into smaller, more manageable operations that can be executed efficiently.

The system applies partial aggregation by focusing computational resources on regions where overlaps actually occur, rather than processing the entire image uniformly. By identifying and aggregating only the relevant overlapping regions of interest, the system achieves improved detection accuracy without unnecessarily increasing processing complexity across the entire image data.

The patent performs preliminary aggregation of confidence scores before generating heatmaps. By pre-computing the aggregate confidence values for all overlapping regions and storing them in an optimized data structure, the system prepares the data in advance for heatmap generation. This preliminary action reduces the computational burden during the actual heatmap creation process, as the aggregation work has already been completed.

The system uses the aggregated confidence scores to automatically generate heatmaps without requiring additional manual processing or intervention. The heatmap generation process leverages the pre-computed aggregate scores directly, allowing the system to serve its own localization needs efficiently. This self-service approach minimizes additional computational requirements by reusing already-processed data.