Lightweight night target detection method and system based on deep learning

By building a lightweight night object detection model based on deep learning, the problem of reduced accuracy and reliability in night object detection is solved, and accurate object detection and computing resources are achieved in low-light environments.

CN119723045APending Publication Date: 2025-03-28NANJING UNIV OF SCI & TECH
4 Cites 6 Cited by

Patent Information

Application Number
CN202411789501.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-06
Publication Date
2025-03-28

AI Technical Summary

Technical Problem

The prior art has problems of reduced accuracy and reliability in night object detection, especially in low-light environments, the generalization ability of traditional object detection algorithms is poor and the computing resource consumption is large, making it difficult to achieve lightweight object detection.

Method used

A lightweight night object detection method based on deep learning is adopted, and a lightweight night object detection model is built through the image enhancement module, feature extraction module, feature fusion module, hybrid channel attention module and result output module to reduce the calculation complexity and improve the detection accuracy.

Benefits of technology

Accurate object detection is achieved in low-light environments, improving detection accuracy and reliability, while reducing the consumption of computing resources and meeting the needs of lightweight.

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

Abstract

The invention discloses a light-weight night target detection method and system based on deep learning. The method comprises the following steps: acquiring a night target image to be detected; inputting the obtained night target image into a pre-constructed night target detection model, sequentially performing image enhancement, feature extraction and feature fusion of different levels, and outputting a detection result; the night target detection model comprises an image enhancement module used for carrying out image enhancement and exposure correction on an input image to obtain an enhanced target image; the feature extraction module is used for performing feature extraction on the enhanced target image obtained by the image enhancement module to obtain feature maps of different scales; the feature fusion module is used for fusing the feature maps of different scales output by the feature extraction module; the mixed channel attention module is applied to image fusion of the feature fusion module by combining local and global features and information of channel and spatial features; and the result output module is used for processing the fused feature map and outputting a detection result of the model. According to the invention, the backbone network is improved through image enhancement, introduction of a mixed channel attention mechanism and adoption of a lightweight downsampling module, so that the accuracy of target detection is improved while network lightweight is realized.
Need to check novelty before this filing date? Find Prior Art