Unmanned aerial vehicle small target detection method and system for few samples and storage medium thereof

By combining artificial feature extraction and deep learning and introducing an adaptive feature weight adjustment mechanism, the problem of scarcity of data in small-sample detection is solved, and the detection performance and robustness are significantly improved.

CN119693608APending Publication Date: 2025-03-25CHENGDU HAOFU TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202411617584.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-13
Publication Date
2025-03-25

AI Technical Summary

Technical Problem

The existing small-sample detection methods rely highly on a large amount of labeled data, resulting in limited data samples in actual operations, making it difficult to effectively capture the characteristics of small-sample, affecting the accuracy and reliability of detection.

Method used

Using a method combining manual feature extraction and deep learning, we integrate features such as shape, texture, and color with features automatically extracted by deep learning models, and introduce an adaptive feature weight adjustment mechanism to dynamically adjust the weight of the feature to improve the effectiveness of the fusion feature.

Benefits of technology

It significantly improves the target detection performance under the condition of small samples, enhances the model's learning ability under a small amount of data, improves the accuracy and robustness of detection, and ensures that efficient and accurate target detection capabilities are still maintained in a variety of environments.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses an unmanned aerial vehicle small target detection method and system for few samples and a storage medium thereof. The method comprises the following steps: preprocessing image data acquired by an unmanned aerial vehicle; manually performing feature extraction on the preprocessed image data, wherein the extracted features comprise shape features, texture features, color features and key point and local features; fusing the extracted features, and forming a feature vector; carrying out adaptive feature weight adjustment on the feature vector; and performing target detection according to the adjusted feature vector. According to the method, the precision and robustness of target detection under the condition of few samples can be improved, the learning ability of the model under a small amount of data is enhanced, and the efficient and accurate target detection ability is still kept in a variable environment by improving the effectiveness of the fusion features and the detection performance of the model.
Need to check novelty before this filing date? Find Prior Art