Object Extraction Using Inverted Model Pattern Resizing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object extraction methods in image analysis are inefficient due to the complexity and time-consuming nature of resizing the entire image to match a model pattern, leading to prolonged calculation processes.
Innovation Solution
The method involves resizing a model pattern of an object to be extracted and scanning the resized pattern in a captured image, using forward mapping to map coordinates, which reduces computational complexity and enhances processing speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the whole image is resized to match the model pattern size, then object extraction accuracy is improved, but calculation complexity increases and processing time is prolonged
Solution Approach 1:
Instead of resizing the entire captured image to match the model pattern size, the patent inverts the approach by resizing the model pattern to match various sizes within the captured image. This allows the model to adapt to different object scales without the computational burden of processing the full high-resolution image at multiple scales, thereby maintaining extraction accuracy while reducing calculation complexity
Solution Approach 2:
The patent segments the object extraction process into multiple stages: first identifying candidate regions in the captured image, then applying resized model patterns at different scales to these segmented regions. This segmentation approach avoids the need to process the entire image at full resolution, reducing overall computational complexity while maintaining accuracy in the extracted regions
2Reliability
If the whole image is resized repeatedly to search for objects, then object detection completeness is improved, but processing speed deteriorates
Solution Approach 1:
The patent performs preliminary actions by first identifying candidate regions and estimating object sizes in the captured image before applying the model pattern matching. This preliminary step allows the system to prepare appropriately sized model patterns in advance, avoiding repeated full-image resizing operations and thereby improving processing speed while maintaining detection completeness
Solution Approach 2:
The patent applies partial action by focusing model pattern matching only on identified candidate regions rather than repeatedly processing the entire image. Multiple resized model patterns are applied to these partial regions at different scales, ensuring detection completeness across various object sizes while significantly reducing the total processing area and improving overall processing speed
3Measurement precision
If multiple resized model patterns are used to scan the image, then object extraction accuracy across different scales is improved, but calculation complexity increases
Solution Approach 1:
The patent implements dynamics by creating a set of model patterns with different sizes and scales that can dynamically adapt to objects of varying dimensions in the captured image. Rather than using a single fixed-size model, the system selectively applies appropriately scaled model patterns to candidate regions, improving multi-scale extraction accuracy while managing processing complexity through intelligent pattern selection
Data Source
AI summary
An apparatus and method for quickly and accurately extracting an object in an image, the method including: performing at least one resizing on an extraction model pattern for of the object to be extracted; scanning an image frame, captured by a camera, using each of the at least one resized extraction model pattern, and extracting the object that is matched with any one of the at least one resized extraction model pattern; and displaying information about a position of the extracted object.