Object Detection Using Two-Stage Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection methods in images face a trade-off between precision and speed, where increasing detection speed often compromises precision, and vice versa.
Innovation Solution
A method involving the generation of first and second images using different feature extraction methods, where the first method is faster and less precise, and the second method is slower but more precise, allowing for rapid and precise object detection by classifying images into groups based on Haar, LBP, HOG, SIFT, Gabor, or MCT features, and clustering them for object tracking and recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single precise feature extraction method is used, then detection precision is improved, but detection time increases
Solution Approach 1:
The patent segments the feature extraction process into multiple stages: a first feature extraction method is applied to generate initial candidate images, followed by a second feature extraction method applied only to selected candidates. This segmentation allows the system to achieve both speed (by using the faster first method initially) and precision (by applying the more accurate second method to promising candidates), resolving the contradiction between detection speed and precision.
Solution Approach 2:
The patent applies partial action by using the more time-consuming second feature extraction method only on a subset of images (those that passed the first classification) rather than all images. This partial application of the precise method maintains overall detection speed while ensuring high precision for the final results, effectively resolving the speed-precision trade-off.
2Measurement precision
If multiple feature extraction methods are applied to all images, then detection precision is improved, but processing time increases significantly
Solution Approach 1:
The patent performs preliminary action by applying the first feature extraction method to all images first, creating an initial sorted list of candidates. This preliminary sorting step is then used to guide the subsequent application of the second feature extraction method, ensuring that the more time-consuming precise method is applied only where necessary, thus maintaining high productivity while achieving high precision.
Solution Approach 2:
The patent introduces dynamic adaptability by adjusting which feature extraction method is applied based on the results of the first classification. The system dynamically decides to apply the second method only to images that meet certain criteria from the first method, optimizing the balance between processing speed and precision for each specific image case.
3Speed
If a fast feature extraction method is used, then detection speed is improved, but detection precision deteriorates
Solution Approach 1:
The patent uses the first feature extraction method as an intermediary that provides a preliminary classification and sorting of images. This intermediary step creates a filtered list of candidate images that are then subjected to the second, more precise feature extraction method. The intermediary first method thus enables the system to maintain speed by avoiding the expensive second method on all images, while still achieving precision through its selective application.
Data Source
AI summary
At least one example embodiment discloses a method of detecting an object in an image. The method includes receiving an image, generating first images for performing a first classification operation based on the received image, reviewing first-image features of the first images using a first feature extraction method with first-type features, first classifying at least some of the first images as second images, the classified first images having first-image features matching the first-type features, reviewing second-image features of the second images using a second feature extraction method with second-type features, second classifying at least some of the second images as third images, the classified second images having second-image features matching the second-type features and detecting an object in the received image based on results of the first and second classifying.