Horticultural Image Clustering With Multispectral Optimization And Gaussian Lion Vector Quantization

More and more farmers are turning to unmanned aerial systems (UAS) to keep an eye on their crops with multispectral imagery. To establish UAS-based data as an industry standard for evaluating tree structure, condition, and production parameters, precise data collection and preprocessing protocols are crucial for multi-temporal comparison. This research proposes a novel approach for multispectral horticulture image clustering and optimization making use of methods of deep learning. The multispectral picture that is sent in is subjected to noise reduction and smoothing before being clustered with kernel fuzzy multiscale subspace clustering. The resulting clusters are then optimized with Gaussian lion vector quantization. Experimental evaluation includes accuracy, sensitivity, specificity, Jaccard index, and F-measure metrics using various multispectral image datasets. The proposed method offers a reliable, cost-effective, and efficient way to assess phenotypic traits of citrus varieties and rootstocks, achieving 95% accuracy, 83% sensitivity, 88% specificity, 65% Jaccard index, and 69% F-measure.