文献类型：期刊文献

英文题名：Analysis and positioning of geographic tourism resources based on image processing method with Ra-CGAN modeling

作者：Li, Xiuxia[1]

第一作者：李新鑫

通讯作者：Li, XX[1]

机构：[1]Xinxiang Univ, 191, Jinsui Rd, Xinxiang City, Xinxiang 453003, Henan, Peoples R China

第一机构：新乡学院

通讯机构：[1]corresponding author), Xinxiang Univ, 191, Jinsui Rd, Xinxiang City, Xinxiang 453003, Henan, Peoples R China.|[11071]新乡学院;

年份：2022

卷号：8

期号：4

起止页码：658-668

外文期刊名：AIMS GEOSCIENCES

收录：WOS:【ESCI(收录号:WOS:000863120200001)】；

语种：英文

外文关键词：deep convolutional neural network; remote sensing image segmentation; conditional generative adversarial network (CGAN); attention mechanism

摘要：People's diversified tourism needs provide a broad development space and atmosphere for various tourism forms. The geographic resource information of the tourism unit can vividly highlight the unit's geographic spatial location and reflect the individual's spatial and attribute characteristics. It is not only the main goal of researching the information base of tourism resources, but it is also the difficulty that needs to be solved at present. This paper describes the use of image processing technology to realize the analysis and positioning of geographic tourism resources. Specifically, we propose a conditional generative adversarial network (CGAN) model, Ra-CGAN, with a multi-level channel attention mechanism. First, we built a generative model G with a multi-level channel attention mechanism. By fusing deep semantic and shallow detail information containing the attention mechanism, the network can extract rich contextual information. Second, we constructed a discriminative network D. We improved the segmentation results by correcting the difference between the ground-truth label map and the segmentation map generated by the generative model. Finally, through adversarial training between G and D with conditional constraints, we enabled high-order data distribution features learning to improve the boundary accuracy and smoothness of the segmentation results. In this study, the proposed method was validated on the large-scale remote sensing image object detection datasets DIOR and DOTA. Compared with the existing work, the method proposed in this paper achieves very good performance.