Soil roughness, mainly due to the presence of secondary soil units (such as peds, clods or aggregates), plays an important role in many agricultural phenomena, such as aggregate stability, infiltration, runoff, water stored in depressions, erosion, heat flux and evaporation (Helming et al. 1998; Kamphorst et al. 2000). These processes are essential for soil conservation, affecting germination and growth of seeds, plant available water and eventually influencing fertility and productivity (Sandri et al. 1998). On the other hand, soil macro- and micro-topography are continuously and rapidly evolving, due to the combined impact of many factors, such as wind and rainfall erosion, solar heating, vegetation growth and tillage (Gessesse et al. 2010). Large differences in soil roughness are commonly found not only within the same field (macro-roughness), but also locally within a few meters range (micro-roughness).

1 1178 Open image in new window machine functioning: the performance can be exploited acting on different parameters such as forward speed, mechanical parts positions, pto speed, etc.;

seedbed preparation: to obtain an optimal aggregate size distribution, with ideally no variations due to differences in soil texture, headlands, etc.;

time: the feedback system slows up the operation only when reference parameters exceed given thresholds;

investments: reduced working time leads to reduction of fuel consumption and wear of mechanical parts;

pollution: avoiding excessive soil pulverization and minimizing power overuse, release of pollutants in the atmosphere can be limited; Monitoring cloddiness and roughness variations would make possible the equipping of farm machinery for tillage with an electronic feedback or feed-forward system, as depicted in Fig.. In the feed-forward configuration, the system evaluates the state of the soil ahead of the machine and correspondingly modulates the operation of the working parts, e.g. forward speed, rotating speed of the rotors, tillage depth, angle of tines, discs or deflectors, etc. Conversely, in the feed-back configuration, the instrument carries out a real time analysis on the cloddiness of the soil after machine operations and instantly modifies working parameters, in order to enhance soil finishing homogeneity or keep the soil roughness close to a predetermined optimal value. Such an approach is interesting: indeed taking advantage of the recently introduced ISOBUS communication protocol (ISO3-9:2012) which enables quick and simple data transfer between tractors and implements, it can allow optimization of operations in terms of:

To allow such a “dynamic calibration” approach, instruments are needed for fast three-dimensional characterization of soil roughness. To this end, contact instruments cannot be applied due to their typically slow sampling rate. On the other hand, several non-contact approaches potentially suitable for fast soil sensing have been proposed in the past and in recent years (Viscarra Rossel et al. 2011), mainly based on image analysis techniques (Stafford and Ambler, 1990), laser scanning (Bertuzzi et al. 1990; Darboux and Huang, 2003; Chatzinikos et al. 2013) and photogrammetry (Aguilar et al. 2009). However such techniques suffer from various impediments. Image analysis is not suitable in the case of geometric clod features and may need human intervention for proper segmentation execution (Chimi-Chiadjeu et al. 2014). Laser scanning is a very interesting choice thanks to resolution and fast sampling performances. However, it is better suited to a single profile rather than for whole area analyses and, as costs are still relatively high, it is difficult to justify its implementation in agricultural machines (Draelos et al. 2012). Photogrammetry is also an interesting solution, benefitting from low cost sensors. However, computational time for 3D reconstruction is still an issue.

However, in recent years the advent of faster computer processors is pushing the introduction of new inexpensive sensors, which make the possibility of integration with agricultural machines more realistic.

In the present work, a suitable low-cost sensor based on 3D depth camera technology for dynamic characterization of soil micro-relief is proposed, discussing possibilities and limitations for agricultural applications. The main objectives were to (1) study the performance of the camera under controlled conditions in the laboratory, and (2) study if differences in soil micro-relief due to tillage could be determined under field conditions.