Soil Texture Prediction With Laser Guided Bouyoucos and Support Vector Regression

In this study, with the guidance of a previously presented study, intensity change of the laser beams passed through the soil-water mixture were collected by a microcomputer, and then computerized estimations were made using the Support Vector Regression algorithm. Texture analysis success rate of the method has been compared with traditional hydrometer method. Experiments were made using 10 different soil samples and the magnitude of light signals passing through the soil-water mixture were recorded. Feature extraction was made using the amplitude values of certain periods in the signals and these features were used in the training of a Support Vector Regression system. The obtained sand, silt and clay estimates are presented in comparison with the conventional hydrometer analysis results.