[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2042":3},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"view_count":31,"doi":32,"paper":33,"created_at":49},2042,"An ensemble deep learning approach for soil pH classification leveraging GLCM-based texture features and neural networks","https:\u002F\u002Fdoi.org\u002F10.69598\u002Fsehs.20.26020002","Soil pH-level classification is a critical challenge in precision agriculture owing to data scarcity, high intra-class similarity, and subtle visual variations between pH levels. This research proposed an ensemble-feature deep learning framework that integrates convolutional neural networks with gray-level co-occurrence matrix (GLCM) texture descriptors to address fundamental limitations in pH-level classification. Through comprehensive experiments, we evaluated diverse architectures (i.e., InceptionV3, Inception ResNetV2, ResNet50, and VGG16) enhanced by GLCM features, demonstrating a substantial increase in classification accuracy, particularly under minimal visual distinctions between pH levels. The optimal Inception ResNetV2 w\u002FGLCM configuration achieved the best overall performance, with an F1-score of 82.10%, an accuracy of 82.22%, and a precision of 82.01%, outperforming the conventional machine learning and deep learning approaches evaluated in this study. The proposed method innovatively provides a synergistic integration of hierarchical deep learning features with statistical texture analysis, achieving robust discrimination of visually similar soil sample images capturing different pH levels. These findings advance automated soil analysis and contribute to precision agriculture by establishing a robust solution for image-based soil pH-level classification.","土壤pH值分级是精准农业中的一个关键挑战，其原因在于数据稀缺、类内相似度高以及不同pH值之间细微的视觉差异。本研究提出了一种集成特征深度学习框架，将卷积神经网络与灰度共生矩阵（GLCM）纹理描述符相结合，以解决pH值分级中的根本性局限。通过全面的实验，我们评估了多种经GLCM特征增强的架构（即InceptionV3、Inception ResNetV2、ResNet50和VGG16），结果表明分类准确率显著提升，尤其是在pH值之间视觉差异极小的情况下。最优的Inception ResNetV2结合GLCM配置取得了最佳整体性能，F1分数为82.10%，准确率为82.22%，精确率为82.01%，优于本研究中评估的传统机器学习和深度学习方法。所提出的方法创新性地将层次化深度学习特征与统计纹理分析协同集成，实现了对捕捉不同pH值的视觉相似土壤样本图像的稳健判别。这些发现推进了自动化土壤分析，并通过建立基于图像的土壤pH值分级的稳健解决方案，为精准农业做出了贡献。",null,"Science, Engineering and Health Studies","2026-09-09T00:00:00Z","论文",10,false,72,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},15,20,17,12,8,1,"该研究提出融合GLCM纹理特征与CNN的集成深度学习框架，在土壤pH分级上取得82%以上准确率，方法新颖且对精准农业有实用价值，但属细分技术进展，影响范围有限。",[25],{"name":10,"url":6},[27,28,29,30],"智慧农业","农业人工智能","精准农业","土壤检测",0,"10.69598\u002Fsehs.20.26020002",{"doi":32,"openalex_id":34,"authors":35,"venue":10,"cited_by_count":31,"oa_url":40,"card":41,"direction":47,"ingested_from":48},"W7212037232",[36,38],{"name":37,"orcid":9},"Sujitranan Mungklachaiya",{"name":39,"orcid":9},"Anongporn Salaiwarakul","https:\u002F\u002Fli01.tci-thaijo.org\u002Findex.php\u002Fsehs\u002Farticle\u002Fdownload\u002F266062\u002F182682",{"tldr":42,"method":43,"finding":44,"direction":45,"opportunity":46},"提出融合GLCM纹理特征与CNN的集成深度学习框架，实现土壤pH等级图像分类。","用InceptionV3、ResNet50等CNN结合GLCM纹理特征，在土壤图","Inception ResNetV2+GLCM最优，F1达82.10%、准确率82.22%，优于常规","农业遥感与作物表型","可探索多源数据融合与轻量化模型，提升小样本下土壤属性分类泛化能力。","智慧农业 \u002F 农业物联网","openalex","2026-09-10T23:30:09.461905Z"]