Deformable Object Recognition via Multi-Level Gradient Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems face challenges in efficiently detecting deformable objects under partial occlusion, clutter, and nonlinear transformations, as they often rely on assumptions that sufficiently small parts of the model remain similar, leading to reduced search speed and spurious matches due to the reduced size of parts used in the search process.
Innovation Solution
A holistic approach that transforms the model and search images into a multi-level representation, generates precomputed models with direction vectors, and computes a global match metric by tracking deformation transformations across levels, allowing for robust recognition of objects under general nonlinear deformations without reducing the relevant model size at higher pyramid levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of pyramid levels is increased to improve discrimination of small objects from clutter, then the search accuracy is improved, but the runtime of the object recognition method increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing feature descriptors for all pyramid levels before the actual object recognition search. This allows the system to quickly retrieve and compare pre-processed features during runtime, avoiding the need to compute features on-the-fly for each pyramid level, thus reducing the runtime penalty associated with using multiple pyramid levels while maintaining search accuracy.
2Adaptability or versatility
If the object is transformed by perspective or general deformation, then the adaptability of the recognition system is improved, but the computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the object model into multiple parts or regions, each of which can be independently transformed and matched. This allows the system to handle perspective and general deformations by applying different transformation parameters to different parts of the object, improving adaptability while managing computational complexity through localized processing rather than global transformation of the entire object.
Solution Approach 2:
The patent applies parameter changes by using different transformation parameters (such as perspective transformation matrices or general deformation parameters) to model the object under various viewing conditions. The system adjusts these parameters to match the observed object in the image, enabling robust recognition under deformation while the parameters are optimized to balance accuracy and computational efficiency.
3Productivity
If the model size is reduced at higher pyramid levels to speed up the search, then the search speed is improved, but the discrimination capability between object and clutter is reduced
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing feature descriptors for all pyramid levels before the actual object recognition search. This allows the system to quickly retrieve and compare pre-processed features during runtime, avoiding the need to compute features on-the-fly for each pyramid level, thus reducing the runtime penalty associated with using multiple pyramid levels while maintaining search accuracy.
Data Source
AI summary
The present invention provides a system and method for detecting deformable objects in images even in the presence of partial occlusion, clutter and nonlinear illumination changes. A holistic approach for deformable object detection is disclosed that combines the advantages of a match metric that is based on the normalized gradient direction of the model points, the decomposition of the model into parts and a search method that takes all search results for all parts at the same time into account. Despite the fact that the model is decomposed into sub-parts, the relevant size of the model that is used for the search at the highest pyramid level is not reduced. Hence, the present invention does not suffer the speed limitations of a reduced number of pyramid levels that prior art methods have.