|M.Sc Student||Shlomo Abergel|
|Subject||Quantification of Soil Erosion in Agricultural Fields|
Using Monoscopic Photogrammetry
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Filin Sagi|
|Dr. Naftali Goldshleger|
|Full Thesis text - in Hebrew|
Surface runoff and soil erosion in agricultural fields cause significant damage, and lead to loss of fertile soil and damage to agricultural crops. To quantify erosion and assess the application of the different treatments, means to measure soil loss are needed. Some of the conventional soil quantification methods require installation instruments in the field and have an indirect impact on the erosion. Others are cumbersome to apply and may lead to crude or erroneous measurements. Additionally, they are predominantly based on manual work in the field and on a subsequent lengthy and error prone processing stage back in the office.
The research presents a photogrammetric mensuration method that offers high level of automation in calculating the surface volume. The model is based on terrestrial monoscopic photogrammetry for autonomous estimation of volumes and the consequent estimation of soil erosion. To support the estimation, the research proposes use of a measuring frame that provides both a means for georeferencing and the definition of the profile plane in object space. A piece of fabric that is laid on the ground defines the actual profile shape.
Two different strategies are studied for the profile area estimation. The first considers the radiometric content in the image and seeks means that allow detection of the measuring frame, alignment of the image, and finally detection of the profile and estimation of its shape and area. To characterize the profile, a set of steerable filters and robust curve fitting algorithms provide a reliable result which is robust to the furrow's geometry, shadows, variations in illumination, and dirt.
The second approach follows the same general steps but attempts exploiting information existing in color space to facilitate a more immediate processing of the images. The proposed model makes use of the sRGB space properties as derived from the chromaticity diagram to make the analysis robust to variations in illumination, shadow effect, and imaging. A proposal of a new scheme that allows color consistency under varying illumination and shadowing condition is then proposed.
Both solutions perform well under varying imaging and illumination conditions. They enable configurational analysis of the variables in raised bed profiles as a basis for analyzing erosion. These results show that the proposed methods enable soil researches to optimize the usage of soil conservation and to calibrate existing models, and provide valuable assessment tool when establishing new ones.