TOF Range Image Distance Error Correction via Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Range image sensors, particularly those using the time of flight method, face accuracy issues due to varying distance measurement errors caused by installation environments and temperature changes, leading to inaccuracies in detecting the three-dimensional position of a fingertip during gesture operations.
Innovation Solution
An information processing apparatus that repeatedly corrects distance errors in range images by deriving a correction value image based on differences between the range image and a reference image, excluding the hand area, to improve the accuracy of detecting objects in arbitrary positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If distance correction is performed using the entire range image including object areas, then correction processing is simplified, but measurement accuracy deteriorates due to inclusion of erroneous distance data from objects
Solution Approach 1:
The range image is segmented into object areas and non-object areas based on distance information. Correction values are calculated separately for each region, with non-object areas providing accurate reference data while object areas are excluded from correction value calculation. This segmentation resolves the contradiction by enabling accurate correction without including erroneous object data.
Solution Approach 2:
Different regions of the range image are treated with different correction strategies. Non-object areas undergo correction value calculation while object areas are identified and excluded. This local differentiation ensures that correction is applied only where valid, maintaining measurement precision without requiring complex global correction processing.
2Measurement precision
If the hand area is excluded from correction value calculation, then measurement accuracy is improved, but device complexity increases due to additional identification processing
Solution Approach 1:
Object area identification is performed as a preliminary step before correction value calculation. By pre-identifying and masking object regions, the subsequent correction processing only needs to handle non-object areas, simplifying the overall workflow while maintaining accuracy. The identification step is performed once and enables efficient batch correction.
Solution Approach 2:
Object areas are extracted and separated from the range image data used for correction calculation. This extraction allows the correction algorithm to focus solely on valid non-object regions, improving accuracy without requiring complex conditional logic throughout the entire correction process. The separated object data can be handled independently or discarded.
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
This approach effectively corrects distance measurement errors in real-time, enhancing the reliability of detecting the three-dimensional position of a fingertip and improving the recognition of touch inputs, such as tap operations, by reducing the influence of measurement errors.
Implementation Method 1
a range image sensor of a time of flight (TOF) method: a target surface is captured using an imaging unit for a range image, from which information on a distance to the subject can be obtained
Data Source
AI summary
An information processing apparatus according to embodiments of the present invention includes an input obtaining unit configured to repeatedly obtain an input image that is obtained through imaging by an imaging unit oriented in a direction intersecting with a predetermined surface, the input image having pixels each representing a distance along the direction, an identifying unit configured to identify, in the obtained input image, an object area where an object that is present between the predetermined surface and the imaging unit is imaged, and a correcting unit configured to correct a distance represented by each pixel in the input image repeatedly obtained by the input obtaining unit, by using a correction value obtained based on a difference between a distance represented by a pixel of a portion of the input image excluding the object area and reference distance between the imaging unit and the predetermined surface.