3D Information Extraction With Statistical Depth Noise Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional information extraction methods using ToF sensors suffer from noise inclusion, leading to unclear boundaries between objects and backgrounds, loss of points of interest, and generation of unwanted noise during smoothing processes, making it difficult to emphasize specific objects and distinguish them from other subjects.
Innovation Solution
A three-dimensional information extraction device and method that includes an image acquiring unit, object extracting unit, depth image acquiring unit, depth value extracting unit, cutout unit, and output unit to extract and emphasize specific objects by calculating statistics on depth values and applying different predetermined ranges based on object outlines and estimated postures to remove noise and enhance object distinction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth information is acquired using a ToF sensor, then distance measurement is achieved, but noise is included in the depth information making boundaries unclear
Solution Approach 1:
The patent extracts only the depth values that fall within the statistical range (mean ± standard deviation) from the depth information, separating valid depth data from noisy outliers. This extraction process removes unreliable depth values while preserving the essential depth measurement capability.
Solution Approach 2:
The patent applies different quality criteria to different depth values based on their statistical properties. By calculating the mean and standard deviation locally for each object region, the system adapts the noise filtering threshold to local depth variations, maintaining boundary clarity while preserving genuine depth features.
2Loss of information
If three-dimensional point group data is generated by combining image and depth information, then 3D information is obtained, but objects and backgrounds cannot be distinguished
Solution Approach 1:
The patent extracts depth values within the statistical range and uses this filtered depth information to generate three-dimensional point group data. By extracting only reliable depth values, the system maintains complete 3D information for valid regions while excluding background noise that would obscure object boundaries.
Solution Approach 2:
The patent changes the parameter selection criterion from fixed threshold filtering to statistical range-based filtering (mean ± standard deviation). This parameter change enables adaptive distinction between objects and backgrounds by using the actual depth distribution characteristics of each scene.
3Reliability
If smoothing process is applied to depth information, then noise is reduced, but unwanted noise is generated and points of interest are lost
Solution Approach 1:
The patent performs preliminary statistical analysis (calculating mean and standard deviation) before generating three-dimensional data, preparing the depth information in advance. This preliminary action identifies valid depth ranges beforehand, preventing loss of points of interest during the subsequent 3D data generation process.
Solution Approach 2:
The patent changes from conventional smoothing operations to statistical range-based filtering. By using mean ± standard deviation as the filtering criterion, the system reduces noise while preserving genuine depth features and points of interest that fall within the statistical range, avoiding the blurring effects of traditional smoothing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method effectively emphasizes specific objects and distinguishes them from other subjects by removing noise and enhancing contrast, allowing for clear three-dimensional information extraction without requiring special imaging environments.
Implementation Method 1
measures a distance to an object by measuring a time required from emission of light from the light emitting element to reception of light by the ToF sensor through reflection by the object
Data Source
AI summary
A three-dimensional information extraction device includes an image acquiring unit configured to acquire an image in which an object is imaged, an object extracting unit configured to extract an outline of the object included in the acquired image, a depth image acquiring unit configured to acquire a depth image including a plurality of depth values at coordinates in a two-dimensional coordinate system which are distance information for the object, a depth value extracting unit configured to extract three-dimensional information of the object on the basis of the depth values included in the acquired depth image and the extracted outline of the object, a cutout unit configured to cut out the depth values in a predetermined range out of the depth values inside of the extracted outline of the object, and an output unit configured to output image information inside of the extracted outline of the object and the cut-out depth.