Nearest Neighbor Search Using Morton Code Shifting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for nearest neighbor search in point cloud data using Morton codes are inaccurate due to their inability to reflect spatial distance well and periodic jump points, leading to low search accuracy and efficiency, especially when the search range varies.
Innovation Solution
A method that acquires a Morton code set for point cloud data, slices the data based on the code set and a distance threshold, shifts Morton code bits to determine a parent node set, and searches within a predetermined range to find nearest neighbors, with an adaptive approach to adjust the search range based on the density of points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If Morton code is used for nearest neighbour search in point cloud data, then the search process is simplified, but the search accuracy deteriorates because Morton code cannot reflect spatial distance well and has periodic jump points
Solution Approach 1:
The patent introduces an intermediary mapping relationship between Morton codes and spatial coordinates. Instead of directly using Morton code proximity to determine spatial proximity, the system establishes a mapping that allows translation between Morton code space and actual spatial space, thereby resolving the contradiction between operational simplicity and measurement precision
Solution Approach 2:
The patent changes the parameter representation by introducing multiple levels of Morton code interpretation. It uses the same Morton code system but interprets it at different granularities and with different reference frames, allowing the system to maintain operational simplicity while achieving accurate spatial distance reflection through parameter transformation
2Productivity
If a fixed search range is used around Morton code, then the search efficiency is maintained, but the adaptability to different point densities deteriorates
Solution Approach 1:
The patent makes the search range dynamic by adjusting it according to the local density of points. Instead of using a fixed search range, the system adapts the search range based on the distribution characteristics of points in different regions, thereby maintaining high search efficiency in sparse areas while ensuring adequate coverage in dense areas
Solution Approach 2:
The patent applies different search strategies to different local regions of the point cloud. By analyzing local point density and applying appropriate search ranges or methods for each region, the system achieves both efficiency and adaptability, with each local area receiving optimized treatment according to its specific characteristics
3Measurement precision
If the search range is increased to improve accuracy, then the nearest neighbour search accuracy improves, but the computational complexity and time consumption increase
Solution Approach 1:
The patent segments the search space into multiple regions based on Morton code hierarchy and spatial partitioning. Instead of searching the entire point cloud or a large fixed range, the system divides the search into smaller, manageable segments that can be processed independently, thereby maintaining high accuracy while reducing overall computational complexity
Solution Approach 2:
The patent performs preliminary organization of point cloud data using Morton codes before the actual nearest neighbour search. By pre-sorting and structuring the data according to Morton code sequences, the system creates an optimized search foundation that enables accurate results to be obtained with reduced computational effort during the actual search phase
Data Source
AI summary
A nearest neighbour search method, an encoder, a decoder and a storage medium are provided. The method includes: acquiring a Morton code set of point cloud data to be searched; according to the Morton code set and a first distance threshold value, layering the point cloud data to obtain current-layer data; shifting, by a first preset number of bits, to the right a Morton code bit of prediction data corresponding to the current-layer data in the point cloud data, so as to obtain a corresponding first parent node set; according to a Morton code of the current-layer data, determining a neighbour area in the first parent node set, and determining, within the neighbour area, a nearest neighbour point set of the current-layer data and performing searching within a first preset search range corresponding to the search point to determine the nearest neighbour point set.