Scholarlink

<p>Leveraging information from spatial, spectral, and temporal domains is crucial for enhancing classification results. To effectively integrate these diverse sources, the present research aimed to investigate the combined effects of the contextual Possibilistic <i>c-</i>Means with Constraints (PCM-S) model and the Individual Sample as Mean (ISM) training approach. The study utilized the synergistic capability of three distinct components, data, model, and training approach, to produce an enhanced mapping of harvested paddy fields near Patiala City, Punjab. By incorporating the temporal phenological stages of paddy, the harvested paddy fields were effectively distinguished. The local convolution model reduced noise by incorporating spatial constraints and yielded accurate classification results, while the ISM training approach handled heterogeneity within the harvested paddy fields. Additionally, the trend of paddy harvesting was determined from the area and percentage of date-wise harvesting. Notably, among 155.42 sq. km of paddy fields in the vicinity of Patiala, 20.74% of the total paddy area underwent harvesting across thirteen mapped dates, constituting an area of 32.23 sq. km. The accuracy assessment results indicated a minimum intra-class Mean Membership Difference (MMD) (nearly 0), maximum inter-class MMD (almost 1), lower variance and entropy, and F-Score exceeding 0.8. These findings underscore the reliability of the results.</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

Optimizing Harvested Paddy Field Classification: Leveraging Combined Local Convolution and Individual Sample as Mean Training Approach

  • 期刊论文

  • 发表日期：

    2025年8月14日

查看全文
  • Anamika Palavesam Sarathamani,
  • Anil Kumar,
  • Raghavendra Pratap Singh

摘要

Leveraging information from spatial, spectral, and temporal domains is crucial for enhancing classification results. To effectively integrate these diverse sources, the present research aimed to investigate the combined effects of the contextual Possibilistic c-Means with Constraints (PCM-S) model and the Individual Sample as Mean (ISM) training approach. The study utilized the synergistic capability of three distinct components, data, model, and training approach, to produce an enhanced mapping of harvested paddy fields near Patiala City, Punjab. By incorporating the temporal phenological stages of paddy, the harvested paddy fields were effectively distinguished. The local convolution model reduced noise by incorporating spatial constraints and yielded accurate classification results, while the ISM training approach handled heterogeneity within the harvested paddy fields. Additionally, the trend of paddy harvesting was determined from the area and percentage of date-wise harvesting. Notably, among 155.42 sq. km of paddy fields in the vicinity of Patiala, 20.74% of the total paddy area underwent harvesting across thirteen mapped dates, constituting an area of 32.23 sq. km. The accuracy assessment results indicated a minimum intra-class Mean Membership Difference (MMD) (nearly 0), maximum inter-class MMD (almost 1), lower variance and entropy, and F-Score exceeding 0.8. These findings underscore the reliability of the results.