Site-specific weed control technologies

DOI: 10.1111/j.1365-3180.2009.00696.x

Site-specific weed control technologies S CHRISTENSEN*, H T SØGAARD , P KUDSKà, M NØRREMARK§, I LUND–, E S NADIMI– & R JØRGENSEN– *Department of Agriculture and Ecology, Faculty of Life Sciences, University of Copenhagen, Taastrup, Denmark,  Engineering College of Aarhus, A˚rhus C, Denmark, àDepartment of Integrated Pest Management, Faculty of Agricultural Sciences, University of Aarhus, Flakkebjerg, Slagelse, Denmark, §Department of Agricultural Engineering, Faculty of Agricultural Sciences, University of Aarhus, Tjele, Denmark, and –Faculty of Engineering, University of Southern Denmark, Odense, Denmark Received 16 August 2008 Revised version accepted 3 November 2008

Summary Site-specific weed control technologies are defined as machinery or equipment embedded with technologies that detect weeds growing in a crop and, taking into account predefined factors such as economics, take action to maximise the chances of successfully controlling them. In this study, we describe the basic parts of site-specific weed control technologies, comprising weed sensing systems, weed management models and precision weed control implements. A review of state-of-the-art technologies shows that several weed sensing systems and precision implements have been developed over the last two decades, although barriers prevent their break-

through. Most important among these is the lack of a truly robust weed recognition method, owing to mutual shading among plants and limitations in the capacity of highly accurate spraying and weeding apparatus. Another barrier is the lack of knowledge about the economic and environmental potential for increasing the resolution of weed control. The integration of site-specific information on weed distribution, weed species composition and density and the effect on crop yield, is decisive for successful site-specific weed management. Keywords: weed patches, weed management, weed sensing, weed recognition, patch spraying, precision weeding.

CHRISTENSEN S, SØGAARD HT, KUDSK P, NØRREMARK M, LUND I, NADIMI ES & JØRGENSEN R (2009). Site-specific weed control technologies. Weed Research 49, 233–241.

Introduction Oerke (2006) estimated the potential crop yield loss without weed control at 43%, on a global scale. Most weeds are either controlled mechanically, by some form of cultivation or chemically, by application of herbicides. Of the vast tonnage of chemical herbicides applied, a large proportion is lost because of drift or evaporation, deposited on the crop or the soil and only a low percentage of the herbicide reaches the target weeds. Besides having potentially adverse environmental impacts and giving rise to concerns over the possible effects on human health of pesticide residues in food and drinking water, herbicides and their application represent a significant variable cost in crop production. These concerns have led to legal

regulations covering herbicide usage in several countries and an increasing demand for organic foodstuffs produced without the usage of herbicide. Therefore, an essential part of the progress toward economically and environmentally sustainable weed management is new weed control technology. The spatial heterogeneity of weeds has inspired several weed scientists to study the species distribution of the plants (Wiles et al., 1992; Heisel et al., 1996; Rew & Cousens, 2001; Gonzalez-Andujar & Saavedra, 2003) and different technologies for weed detection, spatial weed management and spatial variable application of herbicides (Gerhards et al., 1997; Christensen & Heisel, 1998; Paice et al., 1998; Gerhards & Oebel, 2006). Although the studies and experiments have shown

Correspondence: S Christensen, Department of Agriculture and Ecology, Faculty of Life Sciences, University of Copenhagen, Højbakkega˚rd Alle´ 30, DK-2630 Taastrup, Denmark. Tel: (+45) 3058 9614; Fax: (+45) 3533 2175; E-mail: [email protected]  2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

234 S Christensen et al.

significant potential savings and technical progress in sensing, weeding and spraying technologies, few of these technologies have been commercially exploited. During the last decade, a broad and rapidly expanding range of new technologies for precision agriculture has been developed and implemented in agricultural practices, along with low-cost geo-positioning services from satellite systems, such as GPS. The range of technologies can be categorised into three types of hardware and software. The first category is the crop and yield monitors and related mapping software, e.g. Yara N-sensor (Yara International ASA, Norway), GreenStar (Deere & Company, IL, USA), AgLeader (Ag Leader Technology, Inc., IA, USA) and PFadvantage (Ag Leader Technology). The second category of precision agricultural technologies uses geographical information systems (GIS) to manage the large amount of data (field boundaries, soil parameters, yield data and other field data from previous years) and tools for deciding and planning the spatially variable application of nutrients and pesticides. This second group includes the hardware and software FarmWorks (Farm Works Software, IN, USA), AgroGuide (Deere & Company) and AGROffice (PROGIS Software GmbH, Austria). The third category of hardware and software that has appeared within the last five years concerns the automatic steering and guidance of vehicles and implements, e.g. the systems AutoFarm (Novariant, Inc., California, USA), TruPath (Topcon-Sauer/Danfoss joint venture company, Neumunster, Germany), AutoPilot (Trimble Navigation Limited, CO, USA) and AutoTrac (Deere & Company). Some of the commercialised hardware and software comprises tools for weed mapping and control software that adapts spraying to sites of local weed occurrence. Few farmers, however, have adopted site-specific weed management, although several studies have shown that weed occurrence and density varies significantly within a farm or a field (Lutman & Miller, 2007). Gerhards and Christensen (2003) reviewed real-time weed sensing, decision-making and patch spraying in maize, sugar beet, winter wheat and winter barley. The authors found that the main barrier was the balance between the potential savings and the cost of weed sensing. The cost of manual weed sensing was too high and it was concluded that automated weed sensing systems were a prerequisite for further progress in site-specific weed management. Another, perhaps more decisive, reason is that the technologies developed so far are dedicated to specific crops and ranges of weed species. Sensing a large, unknown number of species, while simultaneously making instantaneous decisions about the level of control, choice of herbicide etc, is still a very complex process. Dedicated systems therefore limit range of

usage and may increase costs on the farm level, compared with conventional technologies. More generally applicable systems require technologies that can handle the high level of complexity, e.g. using artificial intelligence. Poole et al. (1998) defined artificial intelligence as Ôthe study and design of intelligent agentsÕ, where an intelligent agent is a system that perceives its environment and takes actions that maximise its chances of success. Applying this definition to site-specific weed control technologies, the implement acts autonomously to maximise the chances detecting, selecting measures and controlling weeds in a given crop at a given growth stage according to predefined criteria such as economics. The basic parts of site-specific weed control technologies comprise three key elements: 1 A weed sensing system, identifying, localising and measuring crop and weed parameters. 2 A weed management model, applying knowledge and information about crop–weed competition, population dynamics, biological efficacies of control methods and decision-making algorithms, and optimising treatments according to the density and composition of weed species, economic goals and environmental constraints. 3 A precision weed control implement, e.g. a sprayer with individual controllable boom sections or a series of controllable nozzles that enable spatially variable applications of herbicides. Another essential part of site-specific weed control technologies is related to the perception of the agroecosystem. The traditional hierarchy of ecological systems starts with the individual, increases in level of complexity and expands temporally and spatially to the population, community, ecosystem and the whole biosphere (Kogan & Lattin, 1999). A parallel hierarchy can be identified in agro-ecosystems, which encompass individual crop and weed plants, a small unit of individual plants, cluster or patches of plants within a field, a whole field, a farm with several fields or a regional agro-ecosystem. In terms of weed control, the hierarchy reflected in the spatial resolution within a farm may follow four levels (Fig. 1): 1 Treatment of individual plants with highly accurate spraying nozzles, controllable mechanical implements or laser beams. 2 Treatments of a grid adapted to the resolution, e.g. adjusting the spraying with a nozzle or a hoe unit. 3 Treatment of weed patches or subfields with clusters of weed plants. 4 Uniform treatment of the whole field. Several authors have applied site-specific weed management in fields infested with weeds growing in patches.

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

Site-specific weed control technologies 235

1. Individual plant treatment

2. Treatment of grids

3. Subfield treatment

Fig. 1 The spatial resolution of weed control in a field.

In some of the studies, the effect of spatial treatment resolution has been investigated, but typically the minimum treatment unit considered is 3 by 3 m (Barosso et al., 2004) or 1 by 1 m (Paice et al., 1998). Wallinga (1995) showed that the potential herbicide saving increases with increasing spatial resolution of the weed control. Taking this to its logical conclusion, the greatest saving is achieved when the weed seedlings are treated individually (Søgaard & Lund, 2007). The objective of this study is to review the basic components of site-specific weed control technologies in the context of the four levels of resolution of weed control.

Weed sensing systems Competitiveness and efficiency of herbicides, mechanical and other weed control methods vary significantly among weed species, i.e. it is essential to identify weed species to maximise economy with a minimum environmental impact. A wide range of weed sensing techniques has been studied over the last 10 years. Research progress can be summarised into two categories: aerial-based and ground-based sensing, using digital cameras or non-imaging sensors. With large areas, the most cost-effective approach may be remote sensing, using aircraft or satellites to provide a farm, or a large area encompassing several farms, with maps of weed occurrence. Lamb and Brown (2001) stated that the two requirements for aerial-based remote sensing of weeds were: (i) that suitable differences in spectral reflectance or texture exist between weeds and their background soil and plant canopy and (ii) that the remote sensing instrument has sufficient spatial and spectral resolution to detect weed plants. Brown and Noble (2005) reviewed aerial-based remote sensing and found that these methods can be successfully applied to detect distinct weed patches when the patches are dense and uniform and have unique spectral characteristics, i.e. typically weed patches larger than 1 by 1 m. Aerial-based remote sensing is, therefore, only applicable for whole-field treatments (level 4) or the

4. Whole-field treatment

spatial variable treatment of weed patches or sub-fields with clusters of weed plants (level 3). A major disadvantage of aerial-based remote sensing is that it can be difficult to acquire the data when needed. If weather conditions are not ideal when the satellite or the aircraft passes over, the acquisition can be delayed for days or weeks. Another disadvantage is that it is possible to obtain two similar bulk spectra from two highly different canopy mixtures (Price, 1994). The use of multi-spectral imaging sensors, such as colour digital cameras on a ground-based mobile platform, shows more promise for the spatial treatments at field resolution levels 1, 2 and 3. Greater proximity reduces the pixel sizes to millimetres or smaller, which is a precondition for image analyses of species-specific features, such as shape, texture and plant organisation. With sufficient spatial resolution (below 1 mm), images collected with ground-based camera systems and subsequent image processing routines are able to segment vegetation from soil background and delineate individual weed plants from the crop (Thorp & Tian, 2004). Segmentation, i.e. making the distinction between plant and soil background, is the first step in automated crop and weed sensing. The segmentation may be based on one of the following characteristics of green foliage: 1 The spectral reflectance in the visual spectrum (red, green and blue) (e.g. Woebbecke et al., 1995). 2 The spectral reflectance in the near infrared spectrum (e.g. Hahn & Muir, 1993). 3 A combination of 1 and 2 using bi-spectral cameras built using understanding of the physics of lighting and reflection by vegetation (R, G, near infra red (NIR) wavebands), transmission through band pass filters and reception at the sensor of a CCD camera (e.g. So¨kefeld et al., 2007). 4 Chlorophyll a fluorescence (e.g. Kera¨nen et al., 2003). The second step in automated crop and weed sensing is to distinguish between crops and weed plants. Feyaerts et al. (1999a) described a Low Cost Imaging Spectrograph (LCIS) based on standard optical components. The LCIS has a spectral range from 400 to 1000 nm, a spectral

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

236 S Christensen et al.

resolution of 35 nm, a slit width of 200 lm, and a slit length of 8 mm. The authors compared the classification rates of the LCIS with a conventional spectrograph, with a higher spectral resolution. Under controlled conditions, both systems obtained very high classification rates >85% between 11 classes (one crop and 10 weed species). Pollet et al. (1999) described a system consisting of the LCIS equipped with a CMOS camera (a camera with an integrated circuit containing an array of pixel sensors, each pixel containing a photodetector and an active amplifier), allowing read-out only of spectral lines of interest. Feyaerts et al. (1999b) used the spatial dimension in a tree-based cluster algorithm, making it possible automatically to collect and label training samples for crops emerging in rows. The authors applied a priori knowledge about the row distance to locate the rows and then classify vegetation between the rows as weed. The plant pixels within the rows were classified using the spectral dimension and a dynamic algorithm that continuously adjusted the spectral signatures separating crop and weed according to the prevailing growth and illumination conditions. The algorithm proved to be able to recognise crops and weeds with an accuracy of almost 94%, enabling a significant herbicide reduction, from 15% to 67% (Feyaerts et al., 1999b). Vrindts et al. (2002) used a line imaging spectrograph with higher spectral resolution, in the 420 to 830 nm range, in field conditions with natural light. The obtained correct classification rate of weed species was around 90%, provided the classification model was adapted to the light condition. Kera¨nen et al. (2003) found that chlorophyll a fluorescence alone exhibits species-specific patterns that can be used to distinguish between selected plant species. Fluorescence signals of green vegetation have low intensity in comparison with normal reflection and are thus difficult to measure making it difficult to develop a robust system for outdoor conditions. Kebabian et al. (1999) described a passive system that can measure fluorescence in the oxygen absorption band around 760 nm, where there is no ambient light and no reflection. The performance of this system for weed detection is not yet documented. Morphological characteristics of plant leaves, such as complexity, central moment, principal axis of moment of inertia, first invariant moment, aspect ratio, radius permutation, ratio of perimeter to longest axis, curvature, compactness and elongation, have been used to classify plant species with some success (Lee et al., 1999; A˚strand & Baerveldt, 2002). Tang and Tian (2002) have established a prototype of a system for plant centre measurement, using colour segmentation with dual lookup tables for segmentation of vegetation and plant stem centre, a camera spatial calibration, geometry-

based identification method, a crop row fitting method and manual on-screen correction procedure. The system performed with an accuracy of 95% compared with conventional broadcast application systems. An additional advantage of a system using the DOD technology is that herbicide exposure on the crop and the soil can be avoided. A high-resolution treatment of weeds at level 1 and 2 can also be achieved with highly accurate weeding implements. Examples include mechanical knives that rapidly positioned in and out of the row or a rotating hoe that was lowered to cut the weeds or raised not to damage the crop (Kepner et al., 1978). A˚strand and Baerveldt (2002) utilised a basket weeder rotating in a transverse direction, which could be lowered to uproot weeds. The Ôsmart hoeÕ by Bontsema et al. (1998) is a rotating plate fitted with spring-loaded knives for cutting weeds just above soil surface. When the disc is rotating at 850 rpm, the knives fold out because the centrifugal force is larger than the spring force. When the plant detection system detects a crop plant, the rpm is set to 700 and the knives almost immediately fold in. Wisserodt et al. (1999) developed the Ôcycloid hoeÕ,

which consists of eight rotating tines. The individual, sigmoid-shaped tines can also rotate around their own vertical axis and so stay out of the row, avoiding crop plants. OÕDogherty et al. (2007) have designed a vertical disc that has a cut-out section with peripheral bevels that enable it to avoid the plant stems. To act as a hoe and achieve intra-row control, however, this must operate in the 2–3 cm upper soil level between plants. Diprose and Benson (1984) used a high-voltage (15– 60 kV) electrical discharge to kill single weeds. The method requires that the electrical probe(s) touch the plant or be held in close proximity to it (1–2 cm). Blasco et al. (2002) used an end-effector with a high-voltage electrical discharge probe to kill weeds. Poulsen (2006) described a precision flame weeding system. The system records the position of the crop plants and controls an array of burners on and off when it passes the crop plants. Treatment of single weed plants with a laser beam may be considered as a combination of thermal treatment and a cutting device (Heisel et al., 2001). Mathiassen et al. (2006) carried out an experiment treating apical meristems of the weed species Stellaria media (L.) Vill., Tripleurospermum inodorum (L.) Schultz Bip. and Brassica napus L. with two different types of continuous wave diode lasers. The experiment revealed that laser exposure of the apical meristems at the cotyledon stage has the potential to kill weeds. However, further research is needed to document the efficacy on a broader spectrum of weed species and to improve the precision of the laser application method.

Conclusions Some of the commercial hardware and software developed for precision agriculture practices comprise tools for weed mapping and control software that adapts spraying to local weed occurrence. However, few farmers have adopted site-specific weed management, although several studies have shown that weed occurrence and density varies significantly within a farm or a field. A review of the literature on weed management models showed that the potential economic and environmental effect on crop yield of increasing the resolution of weed control implied the integration of site-specific information about weed species composition, density, emergence, species competitiveness, canopy architecture etc. Species-specific efficacies, for example of different herbicides, were also decisive for the potential saving. An automated weed sensing system that recognises weed species and plants is a prerequisite for saving herbicides for site-specific weed management. A sensing platform may be used to map the weeds and to make a

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

Site-specific weed control technologies 239

treatment map with a weed management model prior to the weed control is carried out. However, several authors have claimed that real-time weed sensing is a precondition for the adoption of site-specific weed management. A wide range of weed sensing techniques has been studied over the last 10 years. So far, none has been developed into a commercial product. It seems that the robustness of the sensing systems, i.e. their ability to cope with the natural variations of spectral or morphological characteristics and mutual shading among weed species in a field requires the application of combinations of high-speed spectral cameras, image processing and embedded algorithms in the weed management model. Several sprayers have been developed for weed control in the field at resolution 3 and 4. Most of the sprayers have systems based on GIS that contain the weed and treatment maps. Another category of precision sprayers is the direct injection sprayers, which allow online adaptation of herbicide type and dosage to the site-specific demand for weed control. These sprayers operate with a series of nozzles, a boom section or the whole boom. Several prototypes of precision implements have been developed for the field resolution 1 and 2. Among them, the application of herbicide droplets from micro-controlled solenoid valves, mechanical shares that rapidly enter and leave crop rows and a rotating hoe or laser beam that cuts the weeds seems to have a significant potential for very accurate control of weeds, e.g. in low competing row crops. Electrical discharge and thermal weed control methods have also been investigated, although these methods may be difficult to control instantaneously.

References ˚ STRAND B & BAERVELDT A-J (2002) An agricultural mobile A robot with vision-based perception for mechanical weed control. Autonomous Robots 13, 21–35. BAROSSO J, FERNANDEZ-QUINTANILLA C, MAXWELL BD & REW LJ (2004) Simulating the effects of weed spatial pattern and resolution of mapping and spraying on economics of sitespecific management. Weed Research 44, 460–468. BERTI A & ZANIN G (1997) GESTINF: a decision model for post-emergence weed management in soybean (Glycine max (L) Merr). Crop Protection 16, 109–116. BLASCO J, ALEIXOS N, ROGER JM, RABATEL G & MOLTO E (2002) Robotic weed control using machine vision. Biosystems Engineering 83, 149–157. BONTSEMA J, VAN ASSELT CJ, LEMPENS PWJ & VAN STRATEN G (1998) Intra-row weed control: a mechatronics approach. In: Proceedings 1st IFAC Workshop Control Applications and Ergonomics in Agriculture (eds N SIGRIMIS & P GROUMPOS), 93–97. International Federation of Automatic Control (IFAC), Athens, Greece.

BRAIN P & COUSENS R (1990) The effect of weed distribution on predictions of yield loss. Journal of Applied Ecology 27, 735–742. BROWN RB & NOBLE SD (2005) Site-specific weed management: sensing requirements – what do we need to see? Weed Science 53, 252–258. CHRISTENSEN S (1994) Crop weed competition and herbicide performance in cereal varieties and species. Weed Research 34, 29–37. CHRISTENSEN S & HEISEL T (1998) Patch spraying using historical, manual and real-time monitoring of weeds in cereals. Zeitschrift fu¨r Pflanzenkrankheiten und Pflanzenschutz Sonderheft XVI, 257–263. CHRISTENSEN S, HEISEL T, WALTER A & GRAGLIA E (2003) A decision algorithm for patch spraying. Weed Research 43, 276–284. COOTES TF, HILL A, TAYLOR CJ & HASLAM J (1994) The use of active shape models for locating structures in medical images. Image and Vision Computing 12, 355–366. DIPROSE MF & BENSON FA (1984) Electrical methods of killing plants. Journal of Agricultural Engineering 30, 197–209. ELLISON AM, DIXON PM & NGAI J (1994) A null model for neighborhood models of plant competitive interations. Oikos 71, 225–230. FELTON WL & MCCLOY KR (1992) Spot spraying. Agricultural Engineering 11, 9–12. FEYAERTS F, POLLET P, VAN GOOL L & WAMBACP P (1999a) Hyperspectral image sensor for weed selective spraying. In: Proceedings of SPIE, Advanced Photonic Sensors and Applications, Vol. 3897 (eds RA LIEBERMAN & AK ASUNDI), 193– 203. Society of Photo-Optical Instrumentation Engineers (SPIE), Singapore. FEYAERTS F, POLLET P, VAN GOOL L & WAMBACP P (1999b) Vision system for weed detection using hyper-spectral imaging, structural field information and unsupervised training sample collection. In: 1999 British Crop Protection Conference, Brighton, UK, 607–614. FROST AR (1990) A pesticide metering system for use on agricultural spraying machines. Journal of Agriculture Engineering Research 46, 55–70. GERHARDS R & CHRISTENSEN S (2003) Real-time weed detection, decision making and patch spraying in maize, sugar beet, winter wheat and winter barley. Weed Research 43, 385–392. GERHARDS R & OEBEL H (2006) Practical experiences with a system for site specific weed control in arable crops using real-time image analysis and GPS-controlled patch spraying. Weed Research 46, 185–193. GERHARDS R, SO¨KEFELD M, SCHULZE-LOHNE K, MORTENSEN DA & KU¨HBAUCH W (1997) Site specific weed control in winter wheat. Journal of Agronomy and Crop Science 178, 219–225. GONZALEZ-ANDUJAR JL & SAAVEDRA M (2003) Spatial distribution of annual grass weed populations in winter cereals. Crop Protection 22, 629–633. HAGUE T, TILLETT ND & WHEELER H (2006) Automated crop and weed monitoring in widely spaced cereals. Precision Agriculture 7, 21–32. HAHN F & MUIR AY (1993) Reflectance Spectra of Brassica Crops and Some Weeds. Department Note, 57, Scottish Centre of Agricultural Engineering (SCAE), Edinburgh, UK, 26 pp.

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

240 S Christensen et al.

HARA T & WYSZOMIRSKI T (1994) Competitive asymmetry reduces spatial effects on size-structure dynamics in plant populations. Annals of Botany 73, 285–297. HEISEL T, ANDERSEN C & ERSBØLL AK (1996) Annual weeds can be mapped with kriging. Weed Research 36, 325–337. HEISEL T, SCHOU J, CHRISTENSEN S & ANDREASEN C (2001) Cutting weeds with a CO2 laser. Weed Research 41, 19–29. KEBABIAN PL, THEISEN AF, KALLELIS S, SCOTT HE & FREEDMAN A (1999) Passive two-band plant fluorescence sensor with applications in precision agriculture. In: Proceedings of SPIE, Precision Agriculture and Biological Quality (eds GE MEYER & JA DESHAZER), 3543, 38–245. Society of PhotoOptical Instrumentation Engineers (SPIE), Bellingham, WA, USA. KEMPENAAR C & LEEMANS KJM (2005) Developments in the selective application of herbicides on pavements. In: Proceedings of the 57th International Symposium on Crop Protection, Ghent, Belgium, 997–1002. KEPNER RA, BAINER R & BARGER EL (1978) Selective mechanical or chemical thinning. In: Principles of Farm Machinery. Section 11.22, 255–257. AVI Publishing company, Inc., Westport, CT, USA. KERA¨NEN M, ARO EM, TYYSTJA¨RVI E & NEVALAINEN O (2003) Automatic plant identification with chlorophyll fluorescence fingerprinting. Precision Agriculture 4, 53–67. KOGAN M & LATTIN JD (1999) Agricultural systems as ecosystems. In: Handbook of Pest Management (ed. JR RUBERSON), 1–35. Dekker Inc, NY, USA. KWON TJ, YOUNG DL, YOUNG FL & BOERBOOM CM (1995) PALWEED: WHEAT: a bioeconomic decision model for postemergence weed management in winter wheat (Triticum aestivum). Weed Science 43, 595–603. LAMB DW & BROWN RB (2001) Remote-sensing and mapping of weeds in crops. Journal of Agricultural Engineering Research 78, 117–125. LAMM RD, SLAUGHTER DC & GILES DK (2002) Precision weed control system for cotton. Transactions of the ASAE 45, 231– 238. LEE WS, SLAUGHTER DC & GILES DK (1999) Robotic weed control system for tomatoes. Precision Agriculture 1, 95–113. LUTMAN P & PCH MILLER (2007) Spatially Variable Herbicide Application Technology; Opportunities for Herbicide Minimisation and Protection of Beneficial Weeds. Research Review No. 62. Home-Grown Cereals Authority, London, UK. Available at: http://www.hgca.com/ (last accessed 25 November 2008). MATHIASSEN SK, BAK T, CHRISTENSEN S & KUDSK P (2006) The effect of laser treatment as a weed control method. Biosystems Engineering 95, 497–505. MORTENSEN DA, JOHNSON GA & YOUNG LJ (1992) Weed distribution in agricultural fields. In: Proceedings of Soil Specific Crop Management, 113–124. ASA-CSSA-SSSA, Madison, WI, USA. NETO JC, MEYER GE & JONES DD (2006) Individual leaf extractions from young canopy images using Gustafson– Kessel clustering and a genetic algorithm. Computers and Electronics in Agriculture 51, 66–85. NORDBO E & CHRISTENSEN S (1995) Spatial variability of weeds. In: Proceedings of the Seminar on Site Specific Farming (ed. SE OLESEN), 67–89. SP Report 26, Nov 1995, Foulum, Denmark.

OÕDOGHERTY MJ, GODWIN RJ, DEDOUSIS AP, BRIGHTON JL & TILLETT ND (2007) A mathematical model of the kinematics of a rotating disc for inter- and intra-row hoeing. Biosystems Engineering 96, 169–179. OERKE EC (2006) Crop losses to pests. Journal of Agricultural Science 144, 31–43. PAICE MER, MILLER PCH & BODLE DJ (1995) An experimental sprayer for the spatially selective application of herbicides. Journal of Agricultural Engineering Research 60, 100–109. PAICE MER, DAY W, REW LJ & HOWARD A (1998) A stochastic simulation model for evaluating the concept of patch spraying. Weed Research 38, 373–388. PEREZ AJ, LOPEZ F, BENLLOCH JV & CHRISTENSEN S (2000) Colour and shape analysis techniques for weed detection in cereal fields. Computers and Electronics in Agriculture 25, 197–212. PETRY W & KU¨HBAUCH W (1989) Automatic distinction of weed species using form parameters by means of digital image processing. Journal of Agronomy and Crop Science 163, 345–351. POLLET P, FEYAERTS F, WAMBACP P & VAN GOOL L (1999) Weed detection based on structural information using an imaging spectrograph. In: Proceedings of 4th International Conference on Precision Agriculture (eds PC ROBERT, RH RUST & WE LARSON), 1579–1591. American Society of Agricultural and Biological Engineers, Precision Agriculture Center, Madison, WI, USA. POOLE D, MACKWORTH A & R GOEBEL (1998) Computational Intelligence: A Logical Approach. Oxford University Press Inc., NY, USA, 576 pp. POULSEN FP (2006) System for selective treatment of plants in row. Patent WO ⁄ 2006 ⁄ 021207. PRICE JC (1994) How unique are spectral signatures. Remote Sensing of Environment 49, 181–186. REW LJ & COUSENS RD (2001) Spatial distribution of weeds in arable crops: are current sampling and analytical methods appropriate? Weed Research 41, 1–18. ROCKWELL AD & AYERS PD (1996) A variable rate, direct nozzle injection field sprayer. Applied Engineering in Agriculture 12, 531–538. RYDAHL P (2004) Crop protection online-weeds – a Danish decision support system for optimisation of herbicide dose rates. In: Proceedings of 54th German Plant Protection Conference, Hamburg, Germany, 249–251. SØGAARD HT (2005) Weed classification by active shape models. Biosystems Engineering 91, 271–281. SØGAARD HT & LUND I (2007) Application accuracy of a machine vision-controlled robotic micro-dosing system. Biosystems Engineering 96, 315–322. SO¨KEFELD M, GERHARDS R, OEBEL H & THERBURG R-D (2007) Image acquisition for weed detection and identification by digital image analysis. In: Proceedings of the 6th European Conference on Precision Agriculture (ECPA), June 3–6, Skiathos, Greece, 523–529. Wageningen Academic Publishers, Wageningen, The Netherlands. STIGLIANI L & RESINA C (1993) SELOMA – expert system for weed management in herbicide-intensive crops. Weed Technology 7, 550–559. SWINTON SM & KING RP (1994) A bioeconomic model for weed management in corn and soybean. Agricultural Systems 44, 313–335.

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241

Site-specific weed control technologies 241

TANG L & TIAN LF (2002) Machine vision for automated corn plant spacing, growth stage and population measurements. Part II: Plant identification. In: Proceedings of the 2002 ASAE Annual International Meeting ⁄ CIGR 15th World Congress. American Society of Agricultural and Biological Engineers, Chicago, IL, USA, Paper No. 023100. THORP KR & TIAN LF (2004) A review on remote sensing of weeds in agriculture. Precision Agriculture 5, 477–508. VRINDTS E, BAERDEMAEKER JDE & RAMON H (2002) Weed detection using canopy reflection. Precision Agriculture 3, 63–80. WALLINGA J (1995) A closer look at the spatial distribution of weeds – perspectives for patch spraying. In: 9th EWRS Symposium: Challenges for Weed Science in a Changing Europe, Budapest, Hungary, 647–653. WEINER J (1982) A neighborhood model of annual-plant interference. Ecology 63, 1237–1241. WILES LJ, OLIVER GW, YORK AC, GOLD HJ & WILKERSON GG (1992) Spatial distribution of broadleaf weeds in North Carolina soybean (Glycine max) fields. Weed Science 40, 554–557.

WILES LJ, KING RP, SCHWEIZER EE, LYBECKER DW & SWINTON SM (1996) GWM: general weed management model. Agricultural Systems 50, 355–376. WILKERSON GG, MODENA SA & COBLE HD (1991) HERB: decision model for post emergence weed control in soybean. Agronomy Journal 83, 413–417. WILSON BJ & BRAIN P (1990) Weed monitoring on a whole farm – patchiness and the stability of distribution of Alopecurus myosuroides over a ten year period. In: Proceedings EWRS Symposium 1990: Integrated Weed Management in Cereals, 45–52. EWRS, Vammela, Finland. WISSERODT E, GRIMM J, KEMPER M et al. (1999) Gesteuerte Hacke zur Beikrautregulierung innerhalb der Reihe von Pflanzenkulturen. In: Proceedings VDI-Tagung Landtechnik Braunschweig, Germany, 155–160. VDI-Verlag, Du¨sseldorf, Germany. WOEBBECKE DM, MEYER GE, VONBARGEN K & MORTENSEN DA (1995) Color indices for weed identification under various soil, residue, and lighting conditions. Transactions of the ASAE 38, 259–269.

 2009 The Authors Journal Compilation  2009 European Weed Research Society Weed Research 49, 233–241