Viewpoint-Adaptive Scale-Down for Object Recognition Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Template matching in object recognition faces challenges with varying scale-down factors affecting resolution and recognition accuracy when objects appear differently from different viewpoints, leading to suboptimal feature extraction and processing speed.
Innovation Solution
An image processing apparatus that determines a scale-down factor for each viewpoint based on the object's three-dimensional shape, generating templates with optimal resolution for each viewpoint to balance recognition accuracy and processing speed, allowing for efficient CPU processing and integration of scale-down factors for batch processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a high scale-down factor is used for viewpoints where the object appears large, then processing speed is improved, but resolution becomes too low and recognition accuracy deteriorates
Solution Approach 1:
The patent applies dynamics by making the scale-down factor adjustable rather than fixed. The system dynamically selects different scale-down factors based on the viewpoint distance, where closer viewpoints (objects appearing large) use smaller scale-down factors to maintain resolution, while farther viewpoints (objects appearing small) use larger scale-down factors to improve processing speed. This dynamic adaptation resolves the contradiction between processing speed and recognition accuracy.
Solution Approach 2:
The patent changes the parameter of scale-down factor according to viewpoint distance. By establishing a correspondence between viewpoint distance and scale-down factor, the system optimizes the balance between resolution and processing speed for each specific viewpoint, thereby resolving the technical contradiction.
2Measurement precision
If a low scale-down factor is used for viewpoints where the object appears small, then recognition accuracy is improved, but processing speed deteriorates due to excessive feature points
Solution Approach 1:
The system dynamically adjusts the scale-down factor based on viewpoint distance. For viewpoints where the object appears small (farther distances), the system uses larger scale-down factors to reduce the number of feature points, thereby improving processing speed while maintaining sufficient recognition accuracy through the dynamic optimization.
Solution Approach 2:
The patent changes the scale-down factor parameter according to the viewpoint distance. By establishing an optimal correspondence relationship, the system achieves the best balance between recognition accuracy and processing speed for each viewpoint condition.
3Device complexity
If a predetermined scale-down factor is used for all viewpoints, then device complexity is reduced, but recognition accuracy varies significantly across different viewpoints
Solution Approach 1:
The patent applies local quality by assigning different scale-down factors to different viewpoint distances rather than using a uniform factor for all viewpoints. This allows each local region (viewpoint distance range) to have an optimized scale-down factor that maintains recognition accuracy, while the overall system remains relatively simple through automated selection.
Solution Approach 2:
The system changes the scale-down factor parameter based on viewpoint distance, establishing an optimal correspondence relationship. This parameter adaptation maintains high recognition accuracy across different viewpoints while avoiding the need for complex manual configuration.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus, an image processing method, an image processing program allow an object recognition process using a scaled-down image with an optimum resolution for each viewpoint. An image processing apparatus (10) includes a scale-down factor determination unit (202) that generates two-dimensional images of an object viewed from a plurality of viewpoints using three-dimensional data representing a three-dimensional shape of the object and determines a scale-down factor for each viewpoint, a template generation unit (203) that scales down the generated two-dimensional image from each viewpoint using the scale-down factor determined for the viewpoint and calculates a feature quantity from the scaled-down image to generate a template, and a template information output unit (204) that outputs, for each viewpoint, the template and the scale-down factor used for generating the template in a manner associated with each other.