Dual-Direction Feature Extraction for Inclined Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image object recognition methods struggle with accurately recognizing objects that are inclined, rotated, or curved, as they rely on fixed feature information and require multiple object classifiers for different angles, leading to inefficiencies in processing.
Innovation Solution
The method involves extracting lateral and longitudinal feature sequences from images by performing convolution operations in horizontal and vertical angles of view, fusing these features using a weighted sum method, and applying a preset activation function to decode the image feature for recognition, eliminating the need for advance angle determination and multiple classifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple object classifiers are generated for different angles to recognize inclined objects, then recognition accuracy is improved, but processing time increases and system complexity increases
Solution Approach 1:
The patent uses a single object classifier that can process images with objects at any angle by performing feature extraction in both horizontal and vertical directions. This universal approach eliminates the need for multiple angle-specific classifiers, reducing processing time while maintaining recognition accuracy for inclined objects.
Solution Approach 2:
The patent introduces a second dimension of feature extraction by performing convolution operations in both horizontal and vertical directions simultaneously. This dual-directional feature extraction allows the system to capture features from inclined objects without requiring separate classifiers for each angle, thus improving efficiency.
2Measurement precision
If multiple object classifiers are generated for different angles to recognize inclined objects, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The patent employs a single universal object classifier that handles objects at any orientation by extracting features in multiple directions. This reduces device complexity by eliminating the need to generate and manage multiple angle-specific classifiers, while still achieving accurate recognition of inclined objects.
Solution Approach 2:
The patent merges the functionality of multiple angle-specific classifiers into a single classifier by combining horizontal and vertical feature extraction. This consolidation reduces system complexity while maintaining the ability to recognize objects at various angles accurately.
3Productivity
If feature information is specified in advance for fixed posture objects, then processing efficiency is improved, but adaptability to inclined or deformed objects deteriorates
Solution Approach 1:
The patent makes the feature extraction process dynamic by performing convolutions in both horizontal and vertical directions based on the image content. This dynamic approach allows the system to adapt to objects at any angle or deformation state while maintaining processing efficiency, unlike static pre-specified feature information.
Solution Approach 2:
The patent changes the parameters of feature extraction by using dual-directional convolutions with different kernels and directions. This allows the system to adapt to various object orientations and deformations without sacrificing processing efficiency, as the feature extraction parameters are adjusted according to the image content rather than being fixed in advance.
Data Source
AI summary
An image object recognition method, apparatus, and computer device are provided. The image object recognition method includes: performing feature extraction in the direction of a horizontal angle of view and in the direction of a vertical angle of view of an image respectively, to extract a lateral feature sequence and a longitudinal feature sequence of the image; fusing the lateral feature sequence and the longitudinal feature sequence to obtain a fused feature; activating the fused feature by using a preset activation function to obtain an image feature; and recognizing an object in the image by decoding the image feature. This solution can improve the efficiency of the object recognition.