Estimating the grape yield before the vintages (the harvest period) is crucial for the winegrower as it allows to: (i) anticipate needed logistics for the harvest, (ii) plan wine commercialization and (iii) adapt production to viticultural regulations (such as AOC). In recent years, proximal (UGV-based or UAV-based) and satellite-based remote sensing instrumentation that relies on optical frequency bands has been used for detecting grapes and estimating the grape yield. However proximal optical sensing techniques exhibit limitations in terms of robustness to environmental conditions and vegetation penetration. To fill the technological gaps, we introduced the use of mm-wave radar sensors that allow to mitigate the limitations of optical proximal sensing. Previous work allowed to demonstrate that it is possible to estimate grape yield with only 1% error using a static radar setup and with 10% error using a UGV-based radar system moving at 0.5m/s. In this thesis, we demonstrate the use of a mobile system (at 1.0m/s) using MIMO radar sensors for the 3D reconstruction of vinerows and the estimation of grape yield from these 3D images. The implemented analysis approaches allowed to estimate grape yield with 13% error on average. We analyzed the sensitivity of the performance of the yield estimators regarding grape variety, UGV trajectory and vibrations and local weather conditions.