Point Cloud Reconstruction via Iterative Reward Function
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud model reconstruction methods face challenges in controlling resolution and efficiently extracting main features, particularly in applications like ultrasonic haptic interfaces, where edge computing devices have limited efficiency and existing filters fail to clearly define resolution, number, and accuracy.
Innovation Solution
A method that randomly selects four non-coplanar reconstruction points and iteratively adds more points based on specific conditions, adjusting the ratio of g to k to optimize the reduction degree of the point cloud model, ensuring each point is selected only once and adhering to rules regarding collinearity and coplanarity, to achieve the highest reduction degree.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If point cloud filters (such as Voxel filter, statistical filter) are used to reduce point number, then the quantity of points is reduced, but users cannot clearly define resolution, number, and accuracy of point cloud structures
Solution Approach 1:
The patent changes the parameters of point cloud reconstruction by introducing a reward function with adjustable weights k and g, and by defining specific reconstruction conditions (total number of points, volume ratio threshold, selection range). This allows users to clearly control resolution, number, and accuracy of the reconstructed point cloud while reducing the point quantity through iterative selection of reconstruction points.
2Productivity
If the number of points in point cloud model is reduced to suit edge computing devices, then computing efficiency is improved, but it becomes difficult to efficiently extract main features with acceptable accuracy
Solution Approach 1:
The patent applies local quality by differentiating between reconstruction points and non-reconstruction points in the point cloud. The iterative selection process identifies and retains key reconstruction points that contribute most to the main features of the point cloud structure, while removing redundant non-reconstruction points. This ensures that the reduced point cloud model maintains acceptable accuracy for feature extraction while improving computing efficiency on edge devices.
3Manufacturing precision
If more reconstruction points are selected to maintain accuracy, then the quality of reconstructed model is improved, but the complexity of the reconstruction process and computational resources increase
Solution Approach 1:
The patent introduces dynamics by making the reconstruction process adaptive through the reward function with adjustable weights k and g. The iterative selection dynamically adjusts which points are chosen as reconstruction points based on the evolving state of the point cloud model and the specific reconstruction conditions. This dynamic approach optimizes the balance between reconstruction accuracy and process complexity, selecting only the necessary number of points rather than using a fixed large set.
Data Source
AI summary
A point cloud model reconstruction method, apparatus, and system are disclosed. In an embodiment, the method includes randomly selecting four non-coplanar reconstruction points in a point cloud model; iteratively selecting other reconstruction points successively until a reconstruction condition is met, and reconstructing the point cloud model based on all reconstruction points. A reduction degree of a reconstructed point cloud model is: Reward=−k·(PointNum−4)+g·VolRate, where PointNum represents a number of current selected reconstruction points, VolRate represents a ratio of a volume of a solid shape, k represents a proportion of the number of the selected points, and g represents a proportion of the volume ratio. The reconstruction method further includes adjusting a ratio of g to k based on user requirements to adjust the reconstruction condition. The mechanism provided in an embodiment can improve resolution of a reconstructed point cloud model and control quality of points in the reconstructed point cloud model.