黑龙江科技大学学报2024，Vol.34Issue(2)：329-334,6.DOI:10.3969/j.issn.2095-7262.2024.02.025
融合机器视觉和无监督域适应的轻型弱小目标检测方法
Light weak target detection method combining machine vision and unsupervised domain adaptation
摘要
Abstract
This paper proposes a light and weak target detection method combining machine vision and unsupervised domain adaptation to address the difficulty of light and weak target detection and track-ing due to small size and weak brightness.The method works by using Gamma correction method to com-pensate for the light and small targets with weak brightness in the image in order to enhance the clarity of the target contour;extracting and fusing feature saliency maps to obtain the target area of the image;training the network by using original image set as the source domain sample,and conducting the target detection with unsupervised domain adaptation by using the target area as the target domain sample through YOLO-V3 network.The results show that the precision accuracy of object extraction improves by 2.05%,and the accuracy of object detection is up to 82.36%with the proposed method,which increa-ses by 2.1% of the accuracy with other methods,as which verifies its better detection effectiveness on the light and weak target detection.关键词
机器视觉/Gamma校正/目标区域提取/YOLO-V3网络/目标检测Key words
machine vision/gamma correction/target area extraction/YOLO-V3 network/object detection分类
信息技术与安全科学引用本文复制引用
武狄,张哲,李强..融合机器视觉和无监督域适应的轻型弱小目标检测方法[J].黑龙江科技大学学报,2024,34(2):329-334,6.基金项目
黑龙江省省属高校基本科研业务费项目(7020000070226) （7020000070226）
