基于图像增强技术的运动目标检测算法
Moving object detection algorithm based on image enhancement technology
李树壮 1祖国明 1李余光 1翟双1
作者信息
- 1. 长春工业大学 计算机科学与工程学院,吉林 长春 130102
- 折叠
摘要
利用多权值联合损失函数进行图像细节恢复来解决图像对比度不高导致的细节模糊问题.针对光流网络在图像特征提取上存在输入参数多的问题,提出空洞残差网络作为图像的特征提取模块,通过增大卷积核的感受野提高特征提取速度.使用 COCO 数据集进行训练和测试.实验结果表明,与已有同类算法对比,文中算法得到的光流场图像有更高的清晰度,有效提高了检测精确度,降低特征提取时间.
Abstract
The multi-weight joint loss function is used to restore image details to solve the problem of blurred details caused by low image contrast;Aiming at the problem that the optical flow network has many input parameters in image feature extraction,a hollow residual network is proposed as the feature extraction module of the image,and the feature extraction speed is improved by increasing the receptive field of the convolution kernel.In this paper,the COCO data set is used for training and testing.The experimental results show that compared with the existing similar algorithms,the optical flow field image obtained by the algorithm in this paper has higher definition,which effectively improves the detection accuracy and reduces the feature extraction time.
关键词
图像增强/目标检测/空洞残差网络/联合损失函数Key words
image enhancement/object detection/empty residual network/joint loss function引用本文复制引用
基金项目
吉林省教育厅项目(JJKH20210739KJ)
出版年
2024
NETL
NSTL
万方数据