3D Dynamic Map Matching for Mobile Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition technologies struggle to accurately identify mobile objects when they are close to or superimposed on targets recorded in three-dimensional environmental map data, leading to recognition difficulties.
Innovation Solution
A system that performs matching between a three-dimensional dynamic map and observation data, identifies candidate points by distinguishing between observation points that do not exist in the dynamic map and those that differ from mapped points, using voxel-based comparisons and monitoring differences within an allowable range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mobile objects are recognized by comparing three-dimensional distance image data and three-dimensional environmental map data, then object recognition can be performed, but recognition accuracy deteriorates when mobile objects are close to or superimposed on targets recorded in the environmental map
Solution Approach 1:
The patent segments the point cloud data into mapping point groups (from environmental map) and observation point groups (from current sensing), then further segments observation points into candidate points that deviate from mapping points. This segmentation enables separate processing of stationary background targets and mobile objects, resolving the contradiction by isolating mobile object detection from the interference of mapped targets.
Solution Approach 2:
The patent extracts candidate points by identifying observation points that do not coincide with mapping points or deviate from their expected positions. This extraction process separates mobile object candidates from the static environmental map background, allowing accurate recognition even when mobile objects are positioned near or superimposed on mapped targets.
2Difficulty of detecting and measuring
If point group data is extracted by comparing three-dimensional distance image data and three-dimensional environmental map data, then object detection can be performed, but detection precision deteriorates due to superposition with map targets
Solution Approach 1:
The patent introduces candidate points as an intermediary concept between raw observation data and final object recognition. By defining candidate points as observation points that deviate from mapping points, the system creates an intermediate representation that filters out static background interference while preserving mobile object signals, thereby improving both detectability and position precision.
Solution Approach 2:
The patent applies local quality by treating different observation points differently based on their relationship with mapping points. Observation points that coincide with or are close to mapping points are treated as background, while those that deviate beyond a threshold are treated as potential mobile objects. This localized differentiation improves detection precision without sacrificing overall detectability.
Data Source
AI summary
By an object recognition device, an object recognition method, or a non-transitory computer-readable storage medium storing an object recognition program, matching between a three-dimensional dynamic map representing a state of a mapping point group obtained by mapping of a target existing in an observation space and three-dimensional observation data representing a state of an observation point group observed in the observation space is performed, a candidate point of the mobile object is searched, and the mobile object is identified.