مرکز منطقه ای اطلاع رساني علوم و فناوري - Determination of weed intensity in wheat production using image processing techniques

Abstract :

It is of great importance to precisely and carefully apply the minimum amount of chemicals as needed because agricultural chemicals negatively impact the human health, environment and balance in the nature and increase the production costs. In this study, it was aimed at determining the density of broad leaf weeds and contributing to the reduction of herbicide use in wheat grown fields. For this purpose, Image Processing Techniques were used in this study; and Artificial Neural Networks (ANN) and regression models were developed for determination of weeds. In the ANN model, Weed Covered Areas Acquired by Image Processing Techniques (WCAAIPT) was used as input parameter; and Actual Weed Covered Areas (AWCA) as output parameter. In the study, a total of 262 data consisting of 244 data for training and 18 data for test were used. In the ANN model, the structure of the network was designed in the form of 1-(9-5)-1, consisting of 1 input layer, 2 hidden layers and 1 output layer; and the number ofneurons in the hidden layer were determined to be 9-5. Also, tansig was used in the first hidden layer, logsig in the second hidden layer; and purelin transfer functions were used in the output layer. In the ANN and Regression models, R^2 value of the ANN model was found to be 99% and the goodness of fit (U^2) to be 0.000436; whereas R^2 and U^2 values of the Regression model were found to be 95% and 0.008431, respectively. It was determined that the results obtained from the ANN model were in agreement with the experimental data. By the developed ANN model, it would be possible to design and manufacture agricultural machinery in order to determine the weed density and reduce the herbicide use.