Weed Mapping Research Articles

A review of 15 years of Spartina management in the San Francisco Estuary

Regional, ecosystem-level conservation projects with significant vegetation management components require planning, coordination, and responsive management strategies to minimize negative impacts and maximize ecological benefits over time. The California State Coastal Conservancy’s Invasive Spartina Project (ISP) offers an example of a complex, ecosystem-scale weed eradication effort guided by regional conservation goals. We review the management framework developed by the ISP, describe decision thresholds used for site-specific management transitions over the project’s 15 years, and present strategies being used to address major challenges to project completion. These strategies include developing genetics and weed mapping approaches to aid with identification of hybrids between the introduced Spartina alterniflora and the native Spartina foliosa. The ISP also developed a tidal marsh restoration project to enhance habitat for an endangered bird, the California Ridgway’s rail (Rallus obsoletus obsoletus), that uses tall, dense forms of hybrid Spartina as high tide refugia and nesting substrate. By 2014, the ISP had installed over 300,000 native plants and recorded a greater than 96 % estuary-wide reduction in hybrid Spartina (from 323 ha to 12 net ha) despite treatment restrictions imposed at 11 sites since 2011 to protect the rail. Approximately 80 % of the remaining hybrid Spartina occurs in areas currently restricted from treatment, delaying project completion. The successes and setbacks of the ISP illustrate the complexities of achieving ecosystem-level conservation goals dependent on large-scale vegetation management.

Sampling for weed spatial distribution mapping need not be adaptive

Weeds are species of interest for ecologists because they are competitors of the crop for resources but they also play an important role in maintaining biodiversity in agroecosystems. To study their spatial distribution at the field scale, only sampled observations are available due to the cost of sampling. Weeds sampling strategies are static. However, in the domain of spatial sampling, adaptive strategies have also been developed with, for some of them, an important on-line or off-line computational cost. In this article we provide answers to the following question: Are the current adaptive sampling methods efficient enough to motivate a wider use in practice when sampling a weed species at a field scale? We provide a comparison of the behaviour of 8 static strategies and 3 adaptive ones on four criteria: density class estimation, map restoration, spatial aggregation estimation, and sampling duration. From two weeds data sets, we estimated six contrasted Markov Random Field (MRF) models of weed density class spatial distribution and a model for sampling duration. The MRF models were then used to compare the strategies on a large set of simulated maps. Our main finding was that there is no clear gain in using adaptive sampling strategies rather than static ones for the three first criteria, and adaptive strategies were associated to longer sampling duration. This conclusion points out that for weed mapping, it is more important to build a good model of spatial distribution, than to propose complex adaptive sampling strategies.

Selecting patterns and features for between- and within- crop-row weed mapping using UAV-imagery

This paper approaches the problem of weed mapping for precision agriculture, using imagery provided by Unmanned Aerial Vehicles (UAVs) from sunflower and maize crops. Precision agriculture referred to weed control is mainly based on the design of early post-emergence site-specific control treatments according to weed coverage, where one of the most important challenges is the spectral similarity of crop and weed pixels in early growth stages. Our work tackles this problem in the context of object-based image analysis (OBIA) by means of supervised machine learning methods combined with pattern and feature selection techniques, devising a strategy for alleviating the user intervention in the system while not compromising the accuracy. This work firstly proposes a method for choosing a set of training patterns via clustering techniques so as to consider a representative set of the whole field data spectrum for the classification method. Furthermore, a feature selection method is used to obtain the best discriminating features from a set of several statistics and measures of different nature. Results from this research show that the proposed method for pattern selection is suitable and leads to the construction of robust sets of data. The exploitation of different statistical, spatial and texture metrics represents a new avenue with huge potential for between and within crop-row weed mapping via UAV-imagery and shows good synergy when complemented with OBIA. Finally, there are some measures (specially those linked to vegetation indexes) that are of great influence for weed mapping in both sunflower and maize crops.

Testing of weed seeking systems for spray trains - development of a test procedure

Introduction G&G is a Hungarian company working as a service provider within the field of vegetation control on railroad tracks. Due to the fact that organic matters can reduce the necessary elasticity of the track installation, weed control is an important issue for the safety of railroad traffic in general. For this reason weed growth in the roadbed and alongside has to be controlled continuously, whereat due to the total size of the railroad network the application of herbicides is the only procedure being economically in our days. From an environmental and also economical point of view the spraying should only take place if weeds really exist within the target area. If not, spraying should be interrupted in order to avoid the disposal of herbicides and to safe them. For this reason G&G has developed a precision spraying system for weed control on railroad tracks, consisting of weed detection and mapping unit as well as a controllable spraying device being able to apply herbicides on nine different segments separately within the target area (Fig. 1). Therefore a project has been initiated with the aim to establish methodologies to determine the quality of the weed detection and mapping system as well as the quality of the precision spraying device related to the target area. The objectives of the project was to determine the • sensitivity and accuracy of the weed detection system with different speeds of the spray train,

A semi-supervised system for weed mapping in sunflower crops using unmanned aerial vehicles and a crop row detection method

This paper presents a system for weed mapping, using imagery provided by unmanned aerial vehicles (UAVs). Weed control in precision agriculture is based on the design of site-specific control treatments according to weed coverage. A key component is precise and timely weed maps, and one of the crucial steps is weed monitoring, by ground sampling or remote detection. Traditional remote platforms, such as piloted planes and satellites, are not suitable for early weed mapping, given their low spatial and temporal resolutions. Nonetheless, the ultra-high spatial resolution provided by UAVs can be an efficient alternative. The proposed method for weed mapping partitions the image and complements the spectral information with other sources of information. Apart from the well-known vegetation indexes, which are commonly used in precision agriculture, a method for crop row detection is proposed. Given that crops are always organised in rows, this kind of information simplifies the separation between weeds and crops. Finally, the system incorporates classification techniques for the characterisation of pixels as crop, soil and weed. Different machine learning paradigms are compared to identify the best performing strategies, including unsupervised, semi-supervised and supervised techniques. The experiments study the effect of the flight altitude and the sensor used. Our results show that an excellent performance is obtained using very few labelled data complemented with unlabelled data (semi-supervised approach), which motivates the use of weed maps to design site-specific weed control strategies just when farmers implement the early post-emergence weed control.

Sugar beet (Beta vulgaris L.) and thistle (Cirsium arvensis L.) discrimination based on field spectral data

Creeping thistle (Cirsium arvensis (L.) Scop.) is a perennial weed that causes yield loss in sugar beet (Beta vulgaris L.) crops. The weeds are usually mapped for site specific weed management because they tend to grow in patches. Remote sensing techniques have shown promising results in species discrimination and therefore provide potential for weed mapping. In this study we examined the feasibility of high-resolution imaging for sugar beet and thistle discrimination and proposed a protocol to select multispectral camera filters. Spectral samples from sugar beet and thistle were acquired with a field portable spectroradiometer under field conditions and Partial Least Squares Discriminant Analysis (PLS-DA) classification models were developed with 211 and 36 spectral features of 1.56 and 10 nm bandwidths, respectively. The classification rates obtained using these models were regarded as the maximum obtainable. Then, spectral responses of a multi-band camera equipped with the filter configuration proposed by the PLS-DA models were simulated. Finally, a simulation of crop-weed discrimination was made using small unmanned aerial vehicles (UAV)-based multispectral images. More than 95% of the thistles and 89% of the sugar beets were correctly classified when continuous spectral data were used with 1.56 and 10 nm bandwidths. Accuracy dropped to 93% of thistles identified and 84% of sugar beets correctly classified when only the four best bands were used. The validation based on aerial images showed that sugar beets and thistle plants could be discriminated in images if sufficient pure pixels containing leaf spectra were available, that is with spatial resolutions of 6 mm pixel−1 or finer.

Early season weed mapping in sunflower using UAV technology: variability of herbicide treatment maps against weed thresholds

Site-specific weed management is defined as the application of customised control treatments only where weeds are located within the crop-field by using adequate herbicide according to weed emergence. The aim of the study was to generate georeferenced weed seedling infestation maps in two sunflower fields by analysing overlapping aerial images of the visible and near-infrared spectrum (using visible or multi-spectral cameras) collected by an unmanned aerial vehicle (UAV) flying at 30 and 60 m altitudes. The main tasks focused on the configuration and evaluation of the UAV and its sensors for image acquisition and ortho-mosaicking, as well as the development of an automatic and robust image analysis procedure for weed seedling mapping used to design a site-specific weed management program. The control strategy was based on seven weed thresholds with 2.5 steps of increasing ratio from 0 % (herbicide must be applied just when there is presence or absence of weed) to 15 % (herbicide applied when weed coverage >15 %). As a first step of the imagery analysis, sunflower rows were correctly matched to the ortho-mosaicked imagery, which allowed accurate image analysis using object-based image analysis [object-based-image-analysis (OBIA) methods]. The OBIA algorithm developed for weed seedling mapping with ortho-mosaicked imagery successfully classified the sunflower-rows with 100 % accuracy in both fields for all flight altitudes and camera types, indicating the computational and analytical robustness of OBIA. Regarding weed discrimination, high accuracies were observed using the multi-spectral camera at any flight altitude, with the highest (approximately 100 %) being those recorded for the 15 % weed threshold, although satisfactory results from 2.5 to 5 % thresholds were also observed, with accuracies higher than 85 % for both field 1 and field 2. The lowest accuracies (ranging from 50 to 60 %) were achieved with the visible camera at all flight altitudes and 0 % weed threshold. Herbicide savings were relevant in both fields, although they were higher in field 2 due to less weed infestation. These herbicide savings varied according to the different scenarios studied. For example, in field 2 and at 30 m flight altitude and using the multi-spectral camera, a range of 23–3 % of the field (i.e., 77 and 97 % of area) could be treated for 0–15 % weed thresholds. The OBIA procedure computed multiple data which permitted calculation of herbicide requirements for timely and site-specific post-emergence weed seedling management.

Spatial Quality Evaluation of Resampled Unmanned Aerial Vehicle-Imagery for Weed Mapping.

Unmanned aerial vehicles (UAVs) combined with different spectral range sensors are an emerging technology for providing early weed maps for optimizing herbicide applications. Considering that weeds, at very early phenological stages, are similar spectrally and in appearance, three major components are relevant: spatial resolution, type of sensor and classification algorithm. Resampling is a technique to create a new version of an image with a different width and/or height in pixels, and it has been used in satellite imagery with different spatial and temporal resolutions. In this paper, the efficiency of resampled-images (RS-images) created from real UAV-images (UAV-images; the UAVs were equipped with two types of sensors, i.e., visible and visible plus near-infrared spectra) captured at different altitudes is examined to test the quality of the RS-image output. The performance of the object-based-image-analysis (OBIA) implemented for the early weed mapping using different weed thresholds was also evaluated. Our results showed that resampling accurately extracted the spectral values from high spatial resolution UAV-images at an altitude of 30 m and the RS-image data at altitudes of 60 and 100 m, was able to provide accurate weed cover and herbicide application maps compared with UAV-images from real flights.

Selection of Grids for Weed Mapping

This study aimed to assess the degree of similarity presented by thematic maps generated by different sampling grids of weed plants in a commercial agricultural area of 7.95 hectares. Monocotyledons and dicotyledons were counted on the 2012/2013 and 2013/2014 harvests, before soybean planting, in the fallow period after wheat harvest, in both years. A regular grid of 10 x 10 m was produced to sample the invasive plants, used as reference, and the counting was done in 1 m² of each sample point, totaling 795 samples in each year, compared to regular grids of 30 and 50 m, generated from the data exclusion of the standard grid. Twenty-two composite soil samples were taken at a depth of 0-20 cm to correlate soil properties with weeds occurrence. For the generation of the thematic maps, the Inverse Distance Weighting (IDW) for interpolation was used; when comparing the maps generated from each grid with the reference map, the kappa coefficient was used to assess the loss of quality of the maps as the number of sample points was reduced. It was observed that the map quality loss was lower in 2013 compared to 2012 when the sampling density of the points was reduced. The 30 x 30 m grids have satisfactorily described the infestation data of the dicotyledons and the 50 x 50 m grids have adequately described the monocotyledon weeds infestation, compared to the standard 10 x 10 m grids.

Derivation of weed intensity maps in wheat (Triticum aestivum) using spatial data with GIS in the central districts of Punjab, India

Remote sensing technologies are playing an increasingly important role in agricultural production. Because of their potential for high spatial and spectral resolution, satellite and aircraft images can contain detailed site specific information about conditions in agricultural fields. It can be used for monitoring crop growth, yield potential, soil conditions, weed intensity etc. For the commercial extension of site-specific herbicide application technology, rapid and cost effective methods for creating accurate weed maps are required. An experiment was conducted to generate weed prescription maps using spatial data with GIS and remote sensing in the central districts of Punjab, India 2007-08 and 2008-2009. The objective of this research was to demonstrate the potential of optical airborne remote sensing in the detection of some specific weeds and their densities in wheat (Triticum aestivum L.) crop in the central districts of Punjab, India. The analysis of spectral and ground measurement was done to select wave bands (wavelength regions) suitable for distinguishing weed-infested and weed-free crop areas. The results of this study showed that weed prescription mapping can be used for forecasting the infestations, on the basis of which farmers can be take the preventive control measures for weed control, which ultimately helps them in preventing yield losses and these weed maps can also be used as an input in yield forecasting models in future.

An Online Replanning Approach for Crop Fields Mapping with Autonomous UAVs

For managing production at the scale of crop fields, maps of plant pests are used to support farmer decisions. Such maps are costly to obtain since they require intensive surveys in the field, most of the time performed by human annotators or with human-controlled Unmanned Aerial Vehicles (UAVs). In this paper, we look at the next challenge from an AI planning point of view: flying fully autonomous UAVs equipped with online sequential decision-making capabilities for pests sampling and mapping in crop fields. Following existing work, we use a Markov Random Field framework to represent knowledge about the uncertain map and its quality, in order to compute an optimised pest-sampling policy. Since this planning problem is Pspace hard, thus too hard to be exactly solved either offline or online, we propose an approach interleaving planning and execution, inspired by recent works on fault-tolerant planning. From past observations at a given time step, we compute a full plan consisting in a sequence of observed locations and expected observations untill the end of the pest-sampling phase. The plan is then applied until the number of actual observations that differ from expected ones exceeds a given threshold, which triggers a new replanning episode. Our planning method favourably compares on the problem of weed map construction against an existing greedy approach - the only one working online - while adding the advantage of being adapted to the autonomous UAVs' flying time constraints.

Quantifying efficacy and limits of unmanned aerial vehicle (UAV) technology for weed seedling detection as affected by sensor resolution.

In order to optimize the application of herbicides in weed-crop systems, accurate and timely weed maps of the crop-field are required. In this context, this investigation quantified the efficacy and limitations of remote images collected with an unmanned aerial vehicle (UAV) for early detection of weed seedlings. The ability to discriminate weeds was significantly affected by the imagery spectral (type of camera), spatial (flight altitude) and temporal (the date of the study) resolutions. The colour-infrared images captured at 40 m and 50 days after sowing (date 2), when plants had 5–6 true leaves, had the highest weed detection accuracy (up to 91%). At this flight altitude, the images captured before date 2 had slightly better results than the images captured later. However, this trend changed in the visible-light images captured at 60 m and higher, which had notably better results on date 3 (57 days after sowing) because of the larger size of the weed plants. Our results showed the requirements on spectral and spatial resolutions needed to generate a suitable weed map early in the growing season, as well as the best moment for the UAV image acquisition, with the ultimate objective of applying site-specific weed management operations.

Weed mapping using techniques of precision agriculture

The aim of this study was to identify and map the weed population in a no-tillage area. Geostatistical techniques were used in the mapping in order to assess this information as a tool for the localized application of herbicides. The area of study is 58.08 hectares wide and was sampled in a fixed square grid (which point spaced 50 m, 232 points) using a GPS receiver. In each point the weeds species and population were analyzed in a square with a 0.25 m2 fixed area. The species Ipomoea grandifolia, Gnaphalium spicatum, Richardia spp. and Emilia sonchifolia have presented no spatial dependence. However, the species Conyza spp., C. echinatus and E. indica have shown a spatial correlation. Among the models tested, the spherical model has shown had a better fit for Conyza spp. and Eleusine indica and the Gaussian model for Cenchrus echinatus. The three species have a clumped spatial distribution. The mapping of weeds can be a tool for localized control, making herbicide use more rational, effective and economical.

Detección de malas hierbas en girasol en fase temprana mediante imágenes tomadas con un vehículo aéreo no tripulado (UAV)

Weed mapping in early season requires of very high spatial resolution images (pixels <5 cm). Currently only Unmanned Aerial Vehicles (UAV) can take such images. The aim of this work was to evaluate the optimal flight altitude for mapping weeds in an early season sunflower field using a low-cost camera that took images in the visible spectrum at several flight altitudes (40, 60, 80 and 100 m). The object based image analysis procedure used for weed mapping was divided in two main phases: 1) crop-row identification, and 2) crop, weed and bare soil classification. The algorithm identified the crop rows with 100% accuracy at every flight altitude (phase 1) and it detected weed-free zones with 100% accuracy in the images captured at 40 and 60 m flight altitude. In weed-infested zones, the classification algorithm obtained the best results in the images captured at low altitude (40 m), reporting 71% of correctly classified sampling frames (phase 2). Most of errors committed (incorrectly classified frames) were produced by non-detection of weeds (negative false). Subsequent studies would consist in a multi-temporal study aiming to detect weeds are at a more advance growth stage. It could reduce the percentage of negative false in the classification.

4WD mobile robot for autonomous steering using single camera based vision system

Purpose – The purpose of this paper is to demonstrate System-of-Systems (SoS) approach to design and development of unmanned robotic platform for greenhouse agricultural application. Design/methodology/approach – SoS design approach is important in developing engineering products. It was observed that while system integration considers designs in a multi-disciplinary level framework, SoS is viewed as a solution focussed approach. In this paper, the authors have demonstrated SoS approach to develop a mobile robot platform. The wheels of the platform are independently controlled by using brushless DC and stepper motors based on fieldbus type Distributed Control System scheme. Findings – The constraints for autonomous traveling were identified during the first phase followed by development of 12 distinct sub-routines during second phase of training. Optimal camera installation angle, driving speeds, steering angle per pixel were found to be valuable constraints for feed-forward parameters for real-time driving. The platform was field tested in a tomato planted greenhouse for yield and weed mapping. Research limitations/implications – The paper focusses on studying vision-based autonomous four-wheel-drive (4WD) constraints and their implementation limitations. Practical implications – The platform was field tested in a tomato planted greenhouse for yield and weed mapping. Social implications – The platform can be used for agricultural operations such as crop scouting, monitoring, spraying, and mapping in a medium to large-scale greenhouse setting. Originality/value – The research and presentation is original. Starting from its mechanical specification to wheel performance study, development of path patterns for training and global navigation algorithm for testing and validation were achieved. The platform can autonomously be driven without any manual intervention.

A Technique for High-Accuracy Ground-Based Continuous Weed Mapping at Field Scale

<abstract><title><italic>Abstract.</italic></title> Site-specific weed management was proven beneficial on both economic and environmental aspects. Little scientific literature exists about weed spatial distribution in corn fields because manual data acquisition is tedious and time-consuming. Airborne and satellite imagery resolution is too low to provide sound data to answer important questions about weed spatial distribution during the critical weed control period (weed seedlings). This article describes a technique to acquire high-resolution ground-based imagery data to quantify weed cover after crop emergence. The technique is based on a mobile platform traveling at a slow walking speed, controlling ambient light, and triggering a camera at fixed intervals. Vegetation is segmented from non-vegetation pixels using principal component analysis, and weed vegetation is distinguished from crop vegetation by location. Automatic image segmentation resulted in more than 99% detection accuracy, and crop-weed distinction resulted in 0.37% error. The technique enabled the creation of 19 one-hectare maps of weed cover with more than 3000 points per map. These high-definition weed maps can be used to analyze the spatial distribution of weeds and the cost-effectiveness of site-specific weed management.

Characterization of weed flora in rubber trees plantations of Bongo (Côte d’Ivoire)

Objectives: the main objective of this study is to characterize the weed floristic diversity of the Bongo rubber trees plantation and to provide a map for sustainable weed managementMethodology and results: a floristic survey of the Para rubber plantations of Bongo (Southeast Côte d’Ivoire) was conducted in 2007 and early 2008. The field tour method was used for weed inventory in 27 plots of 25 ha of different ages. Moreover, the coverage rate of weeds was estimated in samples of 100 square meters each for every plot. Weeds flora is composed of 168 species scattered in the entire rubber trees plantation. These species were distributed in 63 families and 140 genera. The eleven most representative botanical families, constituting more than the half of the flora (50 % of the species), are Rubiaceae, Apocynaceae, Poaceae, Cyperaceae, Euphorbiaceae, Moraceae, Amaranthaceae, Mimosaceae, Fabaceae, Cucurbitaceae and Commelinaceae. Weeds’ mapping was established and illustrated by 4 main weed groups in relation with the stage of rubber trees development. For the plantations of 1 and 2 years old, weeds were dominated by leguminous cover crop Pueraria phaseoloides (Roxb.) Benth. Weed species really begin to establish in the plantation from 8 to 12 years old with appearance of species such as Palisota hirsuta (Thumb.) Engl. and Aneilema beniniense Kunth. Fields of 13 to 23 years old are covered at about 83.5% by Palisota hirsuta (Thumb.) Engl.Conclusion and application of results: The weed mapping will assist in weed managementKey words: weeds, floristic study, planting year, mapping

DEVELOPMENT OF A PROXIMAL MACHINE VISION SYSTEM FOR OFF-SEASON WEED MAPPING IN BROADACRE NO-TILLAGE FALLOWS

Weeds are among the most significant and costly environmental threats in Australian agriculture. Weeds compete with crop plants for moisture, nutrients an d sunlight and can have a detrimental impact on cro p yields and quality if uncontrolled. The distributio n, size, density and species of the weeds are often heterogeneous in the cropping land. Instead of unif ormly spray the same type of herbicide to the whole farm land, selective spray can reduce the herbicide usag e therefore can reduce the serious problems of herb icide resistance, soil damage and food safety. This study describes a weed mapping method which could be used for broadacre no-tillage fallow weed management. The weed maps have the potential to be used as powerf ul herbicide prescription maps for spot spray. The wee d mapping is realized by the machine vision technologies which including image acquisition, ima ge stitching and photomosaic processing. The sampli ng points are continuous and the interpolation methods are used at the minimum levels. The experiment res ult shows that this weed mapping method can map weed under limited conditions.

Study of Optimum Time Span for Distinguishing Rumex spinosus in Wheat Crop Through Spectral Reflectance Characteristics

Using the inherent spatial and temporal variability within a field to manage farm operations is called precision agriculture. This is a site-specific approach which can reduce input costs and results in higher crop productivity, profitability and lesser environmental pollution. Remote sensing provides a means for the development of weed maps by detecting the location of weeds within an agricultural field, so that site-specific/need based herbicide can be applied. Reductions in herbicide use as a result of this practice reduce management costs for farmers and promote environmental friendliness. The results revealed a decreasing trend in the number of tillers, effective tillers, number of grains per ear, 1,000-grain weight and grain yield of wheat with increasing population densities of Rumex spinosus (from 3 to 12 plants m−2). Highest grain yield of wheat (5.75 tonnes ha−1) was recorded under solid stand of wheat and lowest grain yield was recorded in treatments having 12 plants of R. spinosus. Higher radiance ratio and NDVI values were recorded in solid stand or pure wheat treatment and minimum under solid weed plots. It was observed that by using radiance ratio and NDVI, pure wheat can be distinguished from pure populations of R. spinosus after 30 DAS. It remains distinguished upto 120 DAS. Different levels of Rumex populations can be discriminated amongst themselves from 60 DAS onwards.

Reinforcement learning-based design of sampling policies under cost constraints in Markov random fields: Application to weed map reconstruction

Weeds are responsible for yield losses in arable fields, whereas the role of weeds in agro-ecosystem food webs and in providing ecological services has been well established. Innovative weed management policies have to be designed to handle this trade-off between production and regulation services. As a consequence, there has been a growing interest in the study of the spatial distribution of weeds in crops, as a prerequisite to management. Such studies are usually based on maps of weed species. The issues involved in building probabilistic models of spatial processes as well as plausible maps of the process on the basis of models and observed data are frequently encountered and important. As important is the question of designing optimal sampling policies that make it possible to build maps of high probability when the model is known. This optimization problem is more complex to solve than the pure reconstruction problem and cannot generally be solved exactly. A generic approach to spatial sampling for optimizing map construction, based on Markov Random Fields (MRF), is provided and applied to the problem of weed sampling for mapping. MRF offer a powerful representation for reasoning on large sets of random variables in interaction. In the field of spatial statistics, the design of sampling policies has been largely studied in the case of continuous variables, using tools from the geostatistics domain. In the MRF case with finite state space variables, some heuristics have been proposed for the design problem but no universally accepted solution exists, particularly when considering adaptive policies as opposed to static ones. The problem of designing an adaptive sampling policy in an MRF can be formalized as an optimization problem. By combining tools from the fields of Artificial Intelligence (AI) and Computational Statistics, an original algorithm is then proposed for approximate resolution. This generic procedure, referred to as Least-Squares Dynamic Programming (LSDP), combines an approximation of the value of a sampling policy based on a linear regression, the construction of a batch of MRF realizations and a backwards induction algorithm. Based on an empirical comparison of the performance of LSDP with existing one-step-look-ahead sampling heuristics and solutions provided by classical AI algorithms, the following conclusions can be derived: (i) a naïve heuristic consisting of sampling sites where marginals are the most uncertain is already an efficient sampling approach; (ii) LSDP outperforms all the classical approaches we have tested; and (iii) LSDP outperforms the naïve heuristic approach in cases where sampling costs are not uniform over the set of variables or where sampling actions are constrained.