Target Identification Network Model Intra-Inter Class Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target identification methods using center loss functions lack robustness and generalization due to insufficient optimization of inter-class relationships, leading to reduced accuracy in distinguishing between different classes.
Innovation Solution
A target identification method that incorporates both intra-class and inter-class constraints into the loss function, utilizing orthogonal and metric constraints to optimize the relative positions and distances of class centers, thereby enhancing the model's ability to differentiate between classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If center loss function is used for model training, then intra-class distances are optimized, but inter-class relationships are insufficiently optimized leading to reduced robustness
Solution Approach 1:
The patent merges center loss function with triplet loss function to create a composite loss function that simultaneously optimizes intra-class distances (through center loss) and inter-class relationships (through triplet loss), thereby resolving the contradiction between intra-class optimization and model robustness
Solution Approach 2:
The loss function is constructed as a composite of multiple loss components (center loss and triplet loss) with different weighting coefficients, analogous to composite materials, where each component contributes different properties (intra-class compactness and inter-class separability) to achieve overall model robustness
2Manufacturing precision
If center loss function is used for model training, then intra-class compactness is improved, but classification accuracy between different classes deteriorates
Solution Approach 1:
The patent combines center loss (which improves intra-class compactness) with triplet loss (which enhances inter-class separability and classification accuracy) in a unified loss function, allowing both objectives to be achieved simultaneously without compromise
Solution Approach 2:
The patent introduces adjustable weighting coefficients (λ1 and λ2) to dynamically balance the contribution of center loss and triplet loss during training, enabling optimization of both intra-class compactness and classification accuracy through parameter tuning
Data Source
AI summary
A target identification method includes: obtaining an image containing a target to be identified; performing feature extraction on the image to obtain image features in the image; and inputting the image features into a target identification network model to obtain an identification result that determines a class to which the target to be identified belongs. The target identification network model includes a loss function that is based on intra-class constraints and inter-class constraints. The intra-class constraints are to constrain an intra-class distance between sample image features of a sample target and a class center of a class to which the sample target belongs, and the inter-class constraints are to constrain inter-class distances between class centers of different classes, and/or inter-class angles between the class centers of different classes.