Scholarlink

<p>While mapping specific crops grown at different places, the ground truth of every crop site is required. Collecting ground truth at all different sites requires a lot of time and cost. To reduce the ground truth data requirement, a transfer learning approach has been experimented with in this research work. The crop considered for this research work has been Fennel (<i>Foeniculum vulgare</i>), mostly produced in Gujarat and Rajasthan. Fennel is widely cultivated for its edible seeds, leaves, and bulbs used in culinary and medicinal applications. In this research work, the transfer learning approach has been experimented on a fuzzy machine learning model. One fennel site was used to train the fuzzy machine learning model, which was then applied to the untrained fennel sites. To manage heterogeneity in fennel crop fields, “Individual Sample as Mean” (ISM) was employed in this fuzzy MPCM model training strategy. The difficulty in this investigation was determining how well the trained model would function at additional fennel sites in the absence of ground truth data. The transfer learning approach’s primary benefit was a significant reduction in the amount of time needed to train fuzzy machine learning models for crop mapping on a one-time basis. An accuracy of 87% was obtained when the learned fuzzy model was applied to the untrained fennel fields. The methods used for evaluation were Mean Membership Difference (MMD) and Variance for soft output and F-score for hard output. The MMD between the class is 0.9, indicating that there is no spectral overlap and that the training data of fennel, and the other crops are not the same or closer. The MMD within the class was 0.05, indicating that the training and testing belong to the target crop, fennel. The F-score was calculated for the hard output from the report generated from classification and the F-score was 0.8 which is nearly 1 indicating that the classifier performed well, the best overall accuracy is achieved while applying the transfer learning approach using conventional MSAVI2 with the target crop as a training sample.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Study of Transfer Learning Approach in Fuzzy MPCM Model for Fennel Crop Mapping

  • 期刊论文

  • 发表日期：

    2025年12月6日

查看全文
  • Iswarya Muralidharan,
  • Anil Kumar

摘要

While mapping specific crops grown at different places, the ground truth of every crop site is required. Collecting ground truth at all different sites requires a lot of time and cost. To reduce the ground truth data requirement, a transfer learning approach has been experimented with in this research work. The crop considered for this research work has been Fennel (Foeniculum vulgare), mostly produced in Gujarat and Rajasthan. Fennel is widely cultivated for its edible seeds, leaves, and bulbs used in culinary and medicinal applications. In this research work, the transfer learning approach has been experimented on a fuzzy machine learning model. One fennel site was used to train the fuzzy machine learning model, which was then applied to the untrained fennel sites. To manage heterogeneity in fennel crop fields, “Individual Sample as Mean” (ISM) was employed in this fuzzy MPCM model training strategy. The difficulty in this investigation was determining how well the trained model would function at additional fennel sites in the absence of ground truth data. The transfer learning approach’s primary benefit was a significant reduction in the amount of time needed to train fuzzy machine learning models for crop mapping on a one-time basis. An accuracy of 87% was obtained when the learned fuzzy model was applied to the untrained fennel fields. The methods used for evaluation were Mean Membership Difference (MMD) and Variance for soft output and F-score for hard output. The MMD between the class is 0.9, indicating that there is no spectral overlap and that the training data of fennel, and the other crops are not the same or closer. The MMD within the class was 0.05, indicating that the training and testing belong to the target crop, fennel. The F-score was calculated for the hard output from the report generated from classification and the F-score was 0.8 which is nearly 1 indicating that the classifier performed well, the best overall accuracy is achieved while applying the transfer learning approach using conventional MSAVI2 with the target crop as a training sample.