Adaptive Sampling for 3D Pose Estimation

The patent applies local quality by differentiating the treatment of sample points based on their spatial distribution characteristics. Points are classified into dense and sparse regions based on local point density, and different sampling strategies are applied to each region. This allows the system to maintain high precision in error-prone regions while reducing computational load in stable regions, thereby resolving the contradiction between processing speed and measurement precision.

The patent changes the parameter of sample point density adaptively across different regions. By calculating the density of sample points in local neighborhoods and using this information to adjust the number of points to be processed, the system dynamically optimizes the balance between computational efficiency and measurement accuracy. This parameter change enables the system to reduce overall processing time while maintaining precision where it matters most.

The patent segments the set of sample points into multiple groups based on their spatial distribution characteristics. By dividing the points into dense and sparse regions and processing each region with appropriate strategies, the system avoids the need to process all points uniformly. This segmentation enables the system to maintain precision through targeted processing of critical regions while reducing overall processing time through selective processing of less critical regions.

The patent applies partial action by processing only the necessary subset of sample points rather than all points. By identifying and processing points in sparse regions (which have higher influence on precision) while reducing or skipping processing in dense regions, the system achieves sufficient precision with reduced computational effort. This partial action strategy resolves the contradiction by doing just enough processing to maintain accuracy without the excessive action required to process all points.