Detection of Rice Spikelet Flowering for Hybrid Rice Seed Production Using Hyperspectral Technique and Machine Learning

Effective detection of rice spikelet flowering is crucial to the determination optimal pollination timing for hybrid seed production. Currently, status relies on manual observation farmers, which has low efficiency and large errors. This study attempts acquire information using a hyperspectral technique machine learning in order meet needs rapidly automatically. Hyperspectral data male parents with non-flowering two experimental sites were collected an ASD FieldSpec® HandHeld™2 spectrometer. Three traditional classifiers, Random Forest (RF), Support Vector Machine (SVM) Back Propagation (BP) neural network, Convolutional Neural Network (CNN), used build classification models spikelets detection. processing methods, PCA feature extraction, GA selection, combination algorithm, dimensionality reduction. By comparing precision recall rate different algorithms applicable identify investigated. Results show that by evaluating reduction methods model BP extraction. The accuracy reaches up 96–100%. technology algorithm are capable effective flowering. provides technical reference accurate judgment helps determine operation time supplementary rice.