何爽, 卢霞, 张森, 李珊, 唐海童, 郑薇, 林辉, 罗庆龄
江苏海洋大学海洋技术与测绘学院, 江苏 连云港 222005

摘要:
针对传统分类方法易受到"同物异谱"和"同谱异物"影响，致使河口湿地覆盖分类精度较低的问题，提出一种基于遗传算法优化BP神经网络分类算法。以江苏省临洪河口湿地为研究区，选用哨兵Sentinel-2影像，经辐射校正、大气校正和图像裁剪等预处理后，构建基于自适应遗传算法优化的BP神经网络算法开展临洪河口湿地土地覆盖分类研究，并与传统BP神经网络、支持向量机和随机森林算法进行精度比较。研究结果表明：遗传算法优化后的BP神经网络算法开展河口湿地土地覆盖分类的总精度为96.162 7%，Kappa系数为0.952 0；与传统BP神经网络、支持向量机和随机森林分类算法的分类总精度相比，分别提高了7.359 7%、11.677 9%和6.042 4%；对应的Kappa系数也相应提高了0.090 8、0.118 0和0.074 8；有效解决了河口湿地土地覆盖分类精度低的问题。遗传算法优化后的BP神经网络可实现河口湿地土地覆盖的高精度分类，促进湿地资源的合理开发和保护，为实现海洋生态文明建设提供技术支撑。
关键词:河口湿地Sentinel-2土地覆盖分类遗传算法神经网络
DOI：10.11759/hykx20200814002
分类号:TP751
基金项目:国家自然科学基金（41506106）；2019年连云港市“海燕计划”项目；江苏高校优势学科建设工程资助项目（PAPD）；江苏省海洋技术一流专业建设项目；2019年江苏省研究生实践创新项目（SJCX19_0962）；2020年江苏省研究生实践创新项目

Research on classification algorithm of wetland land cover in the Linhong Estuary, Jiangsu Province
HE Shuang, LU Xia, ZHANG Sen, LI Shan, TANG Hai-tong, ZHENG Wei, LIN Hui, LUO Qing-ling
School of Marine Technology and Geomatics, Jiangsu Ocean University, Lianyungang 222005, China
Abstract:
Aiming at the problem that traditional classification methods are susceptible to "same matter with different spectrum" and "same spectrum with foreign matter", resulting in low classification accuracy of estuary wetland coverage, a BP neural network classification algorithm optimized based on genetic algorithm is proposed. The Linhong Estuary wetland in Jiangsu Province was taken as the research area. The Sentinel-2 remote sensing image were chosen and was preprocessed by the radiometric correction, atmospheric correction and image cropping. Due to this, a BP neural network algorithm optimized by adaptive genetic algorithm was conducted to develop the Linhong Estuary wetland land cover classification research. The comparison was performed through classification accuracy among BP neural network algorithm optimized by adaptive genetic algorithm, traditional BP neural network, support vector machine and random forest algorithm. The research results indicated that the total accuracy of BP neural network algorithm optimized by genetic algorithm for estuary wetland land cover classification is 96.162 7%, and the Kappa coefficient is 0.952 0. It was higher than that of the total classification accuracy of traditional BP neural network, support vector machine and random forest classification algorithm. The total accuracy of BP neural network algorithm optimized with adaptive genetic algorithm, traditional BP neural network, support vector machine and random forest algorithm has increased by 7.359 7%, 11.677 9%, and 6.042 4%, respectively. The corresponding Kappa coefficient has also been increased by 0.090 8, 0.118 0, and 0.074 8, respectively. The problem of low accuracy of estuary wetland land cover classification is effectively solved. The BP neural network optimized by genetic algorithm can realize high-precision classification of estuary wetland land cover, promote the rational development and environmental protection of wetland resources, and provide technical support for the construction of marine ecological civilization.
Key words:estuary wetlandSentinel-2land cover classificationgenetic algorithmneural network





PDF全文下载地址:

点我下载PDF