Work

A Moving Object Detection Method Based on Conditional Information and Feature Deepening

Year: 2025

Type: article

Abstract: In the field of moving object detection methods relying on generative adversarial networks, there are problems such as uncontrollable generation results and incomplete extraction of target details. To solve these issues, this paper proposes a moving object detection model that integrates conditional information and feature deepening techniques. First, a multiframe averaging method is designed to derive a clean background model, which serves as a conditional input to guide network optimization with accurate scene information. Then, an autoencoder architecture is constructed by combining multiple small encoders and residual connections, enhancing feature propagation while mitigating detail degradation caused by generative oversampling. Next, a multiscale feature refinement module is devised to further explore the rich semantic information in the deep feature, thereby enhancing the feature representation capability of the network. Finally, experimental results on public datasets demonstrate that the proposed method exhibits stable performance and can achieve better detection results.

Source: Journal of Electrical and Computer Engineering

DOI: 10.1155/jece/5110822

Authors: Hongrui Zhang, Luxia Yang

Institution: Taiyuan Normal University

Language: English

Topic: Video Surveillance and Tracking Methods

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,100

Funder: Shanxi Provincial Key Research and Development Project