SoilMATTic: Arduino-Based Automated Soil Nutrient and pH Level Analyzer using Digital Image Processing and Artificial Neural Network

摘要

In this study, SoilMATTic was developed for faster and accurate soil analysis compared with conventional method to guide farmers on crop suitability and increase farm productivity and crop yield. The Arduino-based prototype automated the whole process of macronutrient and pH analysis of soil from soil testing procedures up to fertilizer recommendation. It includes stepper motors and pumps to fully automate the chemical reaction of soil with chemical reagent during testing and an on board printer to print out fertilizer recommendations. It uses digital image processing technique to efficiently identify (1) Nitrogen, (2) Phosphorus, (3) Potassium and (4) pH level of Philippine farmlands. The system is composed of five stages namely: automated soil testing, image acquisition, image processing, training system, and recommendation. Artificial Neural Network offered fast and accurate performance for the image processing. The system data base stored and manages 356 captured images where 70% is for training, 15% for testing and 15% for validation. Results of this this study showed 96.67 accuracy in identifying soil macronutrient and pH level and gives fertilizer recommendation for Inbred rice plant, Inbred corn, Tobacco, Sugarcane, Pineapple, Mango, Coconut, Abaca, Coffee, Banana through a generated report in printed form.