Development of plant monitoring systems based on multi-camera image processing techniques on hydroponic system

Abstract

The research objective is to develop a monitoring system for the growth of red spinach plants based on image processing techniques from images captured using multiple cameras. The plant used is red spinach (Amaranthusgangeticus L.). Three cameras are installed in the top, side and front position of the plants in the photo box with lighting every 2 days up to 39 days. Model development uses a sample of 236 plants divided into 178 plants used model and 58 plants for model testing every two days. This model tested by the determination coefecient (R2) to measure how much the independent variables ability to explain the dependent variable. The network architecture were three input, first hidden layer (5 neurons), second hidden layer (5 neurons), and output layers with 1 neuron. ANN function with value of the learning level is 0.001. The activation function to predict fresh weight and leaf area of plants is tansig-logsig-tansig and tansig-tansig-logsig. ANN model can predict fresh plant weight with MSE value of 0.02385 and RMSE of 0.154, while for leaf area MSE value of 0.26428 and RMSE of 0.514.