Object Discrimination Using Variation Difference Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object discrimination techniques face challenges in accurately distinguishing objects due to variations in factors like face direction, lighting, and pose, leading to increased complexity and storage requirements for conversion models.
Innovation Solution
An object discriminating apparatus that extracts features from input images, calculates similarity with registered images, derives differences in variations, converts similarity based on these differences, and decides object identity, using a reduced set of conversion models by representing variations as differences rather than combinations, thereby reducing storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conversion models are created for all combinations of object directions and light source directions, then discrimination accuracy is improved, but storage capacity is excessively consumed
Solution Approach 1:
The patent segments the combined variation model into separate component models: one for object direction variations and another for light source direction variations. Instead of storing a single large conversion model for all combinations, the system stores multiple smaller component models that can be independently selected and applied based on the specific variation conditions detected in the input image.
Solution Approach 2:
The patent implements a dynamic model selection mechanism that detects the actual variation conditions (object direction and light source direction) in the input image and automatically selects the appropriate component models. This allows the system to adaptively apply only the necessary models for the current situation rather than using a fixed comprehensive model, optimizing both accuracy and storage efficiency.
2Reliability
If the number of conversion models is increased to cover all variation combinations, then discrimination reliability is improved, but device complexity increases
Solution Approach 1:
The patent divides the complex conversion model into separable component models for different variation factors. This segmentation reduces the overall system complexity by allowing independent management and selection of each component model based on detected variation conditions, rather than managing a single complex model covering all combinations.
Solution Approach 2:
The patent performs preliminary detection of variation conditions (object direction and light source direction) before model application. This preliminary action enables the system to pre-select the appropriate component models, avoiding the need to maintain and process all possible conversion models simultaneously, thereby reducing system complexity while maintaining reliability.
3Speed
If feature quantities are directly compared without normalization, then processing speed is improved, but discrimination accuracy deteriorates due to variation factors
Solution Approach 1:
The patent applies normalization conversion as a preliminary step before feature quantity comparison. By detecting variation conditions and applying the appropriate component models to normalize the input image features beforehand, the system ensures accurate discrimination while maintaining efficient processing, as the normalization is performed using pre-computed conversion models rather than through iterative optimization.
Data Source
AI summary
An object discriminating apparatus is provided with an obtaining unit configured to obtain an input image including an object; an extracting unit configured to extract a feature from the input image; a calculating unit configured to calculate, by collating the feature extracted from the input image and a feature of a previously registered registration image with each other, similarity between the object included in the input image and an object included in the registration image; a deriving unit configured to derive a difference between a variation in the input image and a variation in an output image; a converting unit configured to convert the calculated similarity on the basis of the derived difference between the variations; and a deciding unit configured to decide, on the basis of the converted similarity, whether or not the object included in the input image is identical with the object included in the registration image.