Monocular Endoscope Absolute Dimension Measurement via Autofocus Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional endoscopes, including monocular ones, are limited in measuring absolute dimensions of imaging subjects due to restricted measurement ranges, especially with compact optical systems, and require specialized equipment like stereo optical systems or laser modules, while three-dimensional reconstruction from monocular endoscopes only provides relative dimensions, not absolute ones.
Innovation Solution
An image processing apparatus and method that uses a contrast autofocus endoscope to estimate imaging-subject distances and calculate a scale coefficient for converting relative three-dimensional information to absolute dimensions, employing a learning parameter determined by deep learning on a dataset including images outside the measurement range of the autofocus lens position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a monocular endoscope is used for three-dimensional reconstruction, then device complexity is reduced, but measurement precision of absolute dimensions deteriorates
Solution Approach 1:
The patent introduces an autofocus lens as an intermediary component between the imaging lens and image sensor. By measuring the autofocus lens position and combining it with contrast information, the system can estimate imaging-subject distances and convert relative three-dimensional measurements to absolute dimensions, thereby enabling absolute measurement capability in a monocular system without adding complex stereo optical paths or laser modules
Solution Approach 2:
The patent replaces traditional mechanical measurement systems (stereo optical systems, laser modules) with a computational approach using deep learning estimators. The estimator processes image contrast data and autofocus lens position information to predict absolute distances, substituting physical measurement infrastructure with algorithmic processing while maintaining measurement precision
2Ease of operation
If contrast autofocus method is used, then ease of operation is improved, but measurement range deteriorates
Solution Approach 1:
The patent performs preliminary deep learning training to create an estimator that can predict imaging-subject distances even when autofocus lens positions are at their extremes. By pre-training the model with diverse distance scenarios, the system extends its effective measurement range beyond the mechanical limits of the autofocus lens travel distance
Solution Approach 2:
The patent changes the parameter space by combining multiple measurement parameters (autofocus lens position, image contrast values, estimator predictions) rather than relying solely on autofocus lens position. This multi-parameter approach allows the system to infer distances for subjects beyond the traditional measurement range by using contrast patterns and learned relationships from the training data
3Measurement precision
If specialized equipment like laser module or stereo optical system is added, then measurement precision of absolute dimensions is improved, but device complexity worsens
Solution Approach 1:
The patent makes the monocular endoscope universal by enabling it to perform both standard imaging and absolute dimension measurement functions using the same optical path. The autofocus lens serves dual purposes: focus adjustment for image quality and distance measurement for dimensional analysis, eliminating the need for specialized measurement equipment
Solution Approach 2:
The patent creates a computational model (estimator) that copies the measurement capability of complex systems. By training the estimator on labeled data containing ground truth distances, the model learns to predict absolute dimensions from standard monocular images, effectively copying the functionality of laser-based or stereo-based measurement systems without their hardware complexity
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
Enables the measurement of absolute dimensions of imaging subjects using a general monocular endoscope without specialized equipment, overcoming the limitations of restricted measurement ranges and relative three-dimensional reconstruction, by accurately converting relative dimensions to absolute dimensions.
Implementation Method 1
an endoscope that is capable of automatically adjusting a focal position by means of a contrast autofocus method
Data Source
AI summary
Provided is an image processing apparatus including a processor. The processor is configured to: reconstruct, by employing an image set acquired by means of an endoscope, three-dimensional information of an imaging subject; estimate, by means of an estimator, imaging-subject distances from the image set by employing a learning parameter; calculate, on the basis of the estimated imaging-subject distances and the imaging-subject distances in the three-dimensional information, a scale coefficient; convert relative dimensions of the three-dimensional information to absolute dimensions by employing the scale coefficient; and output the three-dimensional information containing the absolute dimensions. The learning parameter is determined by learning a plurality of learning images including images of imaging-subject distances outside a measurement range in which imaging-subject distances can be measured on the basis of contrast of the image and a position of an autofocus-lens, as well as imaging-subject distances for each of the learning images.