首页|基于人工智能与深度学习的小目标检测方法研究

基于人工智能与深度学习的小目标检测方法研究

扫码查看
为进一步提升图像小目标检测效果,提出一种基于改进YOLOv5的小目标检测方法.其中,以YOLOv5网络作为基本的目标检测方法,通过进行特征层裁剪、引入特征金字塔网络以及快速空间金字塔池化的方式,进一步提升其检测性能.实验结果表明,与未改进的YOLOv5网络相比,改进后的YOLOv5网络具有更好的检测精度,同时轻量化效果良好,检测效率明显提升;与其他目标检测算法相比,基于改进YOLOv5的小目标检测方法能够在保持较高检测精度的同时取得良好的网络轻量化效果,检测效率较高,mAP@0.5为74.98.综上,设计的基于改进YOLOv5的小目标检测方法性能良好,能够进行效果较好的小目标检测,可应用于实际的小目标检测场景中,可行性较高.
Research on Small Object Detection Method Based on Artificial Intelligence and Deep Learning
To further improve the performance of small object detection in images,a small object detection method based on im-proved YOLOv5 is proposed.Among them,the YOLOv5 network is used as the basic object detection method,and the detection per-formance of the network is further improved by pruning its feature layer,introducing a feature pyramid network,and fast spatial pyra-mid pooling.The experimental results show that compared with the unmodified YOLOv5 network,the improved YOLOv5 network has better detection accuracy,good lightweight effect,and significantly improved object detection efficiency;Compared with other object detection algorithms,the small object detection method based on improved YOLOv5 can achieve good network lightweight effect while maintaining high detection accuracy,and has higher detection efficiency,mAP@0.5 It is 74.98.In summary,the designed detection method based on improved YOLOv5 has excellent performance and can perform effective small object detection.It can be applied to practical detection scenarios with high feasibility.

object detectionfeature extractionYOLOv5lightweight design

马文胜、侯锡林

展开 >

辽宁科技大学,辽宁鞍山 114051

目标检测 特征提取 YOLOv5 轻量化设计

2024

自动化与仪器仪表
重庆工业自动化仪表研究所,重庆市自动化与仪器仪表学会

自动化与仪器仪表

CSTPCD
影响因子:0.327
ISSN:1001-9227
年,卷(期):2024.(8)