Leveraging Deep Learning and Farmland Fertility Algorithm for Automated Rice Pest Detection and Classification Model

Seeds will display different germination states during the germination process, and their good or bad state directly influences the subsequent growth and yield of the crop. This study aimed to address the difficulties of obtaining the images of seed germination process in all time series and studying the dynamic evolution law of seed germination state under stress conditions. A licorice sprouting experiment was performed using a seed sprouting phenotype acquisition system to obtain images of the sprouting process of licorice in full-time sequence. A labeled dataset of licorice full-time sequence sprouting process images was constructed based on the four states of unsprouted, sprouted, cracked, and shelled in the sprouting process. An optimized model, YOLOv8-Licorice, was developed based on the YOLOv8-n model and its effectiveness was demonstrated by comparative and ablation tests. Different salt stress environments were simulated via NaCl aqueous solution concentration, and germination experiments of licorice seeds were performed under different salt stresses. The germination state of licorice under different salt stress environments was detected using the YOLOv8-Licorice detection model. Percentage curve of licorice seeds in an unsprouted state displayed a continuous decreasing trend. For the percentage curve of licorice seeds in the sprouted state, an increasing and then decreasing trend was observed under the condition of 0-200 mmol/L NaCl solution, and a continuous increasing trend was observed under the condition of 240-300 mmol/L NaCl solution. Licorice seeds in the cracked state demonstrated percentage curves with an increasing and then decreasing trend under the condition of 0-140 mmol/L NaCl solution and a continuous increasing trend under the condition of 160-300 mmol/L NaCl solution. The percentage curve of licorice seeds in shelled state displayed a continuous increasing trend in 0-200 mmol/L NaCl solution condition and remained horizontal in 220-300 mmol/L NaCl solution condition. Overall, this study provides a valuable method involving the seed sprouting phenotype acquisition system and the proposed method for detecting the germination state of licorice seeds. This method serves as a valuable reference to comprehensively understand the seed sprouting process under triggering treatment.

show abstract

Leveraging Deep Learning and Farmland Fertility Algorithm for Automated Rice Pest Detection and Classification Model

Cited by 4 publications

References 25 publications

Innovative Insect Detection and Classification for the Agricultural Sector Using Gannet Optimization Algorithm With Deep Learning

Innovative Insect Detection and Classification for the Agricultural Sector Using Gannet Optimization Algorithm With Deep Learning

A Deep Learning Model-Based Neural Network Framework for Diabetic Foot Ulcer Classification

YOLOv8-licorice: a lightweight salt-resistance detection method for licorice based on seed germination state

Contact Info

Product

Resources

About