Weeds sampling for map reconstruction: a Markov random field approach

In the past 15 years, there has been a growing interest for the study of the spatial repartition of weeds in crops, mainly because this is a prerequisite to herbicides use reduction. There has been a large variety of statistical methods developped for this problem ([5], [7], [10]). However, one common point of all of these methods is that they are based on in situ collection of data about weeds spatial repartition. A crucial problem is then to choose where, in the eld, data should be collected. Since exhaustive sampling of a eld is too costly, a lot of attention has been paid to the development of spatial sampling methods ([12], [4], [6] [9]). Classical spatial stochastic model of weeds counts are based on Cox processes [3] or kriging [7]. In this work we propose to deal with abundance classes and to adopt a Markov Random Field (MRF) framework. In a companion paper [2], we present an approach for spatial sampling which is based on MRF. This approach relies on an a priori model of the repartition of weeds in crops. It also relies on a model of sampling costs (time spent to sample), in order to mimic eld constraints. The goal of this talk is to present the modelling choices that we have made in order to apply the approach [2] to the sampling and reconstruction problem for a real case study with a large data set of partial samples, in various conditions (weeds, crop, date...). In section 2 we present the model selection study that we have performed in order to build the a priori MRF model of weeds repartition. Then, in section 2, we present the sampling (time) cost model that we have built. Finally, in section 4 we discuss the use of the sampling approach [2] for weeds sampling in crop elds.