ReviewAutomatic GPS-based intra-row weed knife control system for transplanted row crops
Highlights
► Weed control remains a major challenge to agricultural. ► Automatic intra-row weed machine was developed. ► RTK-GPS transplanter plant map and real-time GPS signal was used in actuating the pneumatic knifes. ► Mechanical weed control system for sustainable production of row crops was demonstrated.
Introduction
Mechatronic weed control, specifically the development of an automatic machine for the non-chemical control of weed plants intra-row (within the crop row), remains one of the biggest challenges to agricultural row crop production in industrialized countries today. Intra-row weeds are more difficult to eliminate than inter-row weeds due to their proximity to the crop or seed line. There continues to be an ever increasing interest in the use of mechanical intra-row weeding machines because of the high cost and declining availability of manual labor, concerns over environmental degradation associated with pesticides and an increasing demand for organically produced food (Åstrand and Baerveldt, 2002, Kurstjens, 2007, Dedousis et al., 2007, Tillett et al., 2008, Nørremark et al., 2008).
In geospatial relationship to the crop plants, three weeding zones have been identified: inter-row, intra-row and close-to-crop, Fig. 1 (Griepentrog et al., 2003). Nowadays, mechanical weed control is mainly used and associated with traditional inter-row cultivation. While intra-row hand weeding can be reduced by narrowing the uncultivated band about the seedline through the use of precision inter-row cultivation (Fennimore et al., 2010), most weeds growing in the intra-row region are uncontrolled by inter-row cultivation as are the highly competitive weeds in the close-to-crop zone (Tillett et al., 2002, Melander, 1997). Thus increased research effort is being focused into the development of intra-row weeding systems to remove or destroy the weeds within the row without causing excessive crop damage (Åstrand and Baerveldt, 2002, Bak and Jakobsen, 2004, Blasco et al., 2002, Van Evert et al., 2006, Gobor and Schulze, 2007, Tillett et al., 2008).
Advanced technologies for intra-row weed control have potential for integration and implementation into intelligent systems for arable weed control management strategies. However, finding an overall solution for effective and selective weed control, thus reducing the need for hand weeding, minimizing negative environmental impacts, and increasing economic returns, is not an effortless task, remains great challenges in the mechanization of crop protection. It is critical to maintain a weed-free zone around vegetable crops so the crops do not have to compete with the weeds for water and nutrients.
A major advantage of RTK-GPS mapping technology over machine vision-based methods is that accuracy and precision are independent of the visual appearance of the crop, shadows, missing plants, weed density or other conditions that degrade the performance of machine vision or other plant sensing systems. In addition, no crop specific knowledge, such as visual texture, biological morphology, or spectral reflectance characteristics, is required for operation, simplifying the transition from one crop to another. RTK-GPS auto-guidance based systems can be used to cultivate or spray very close to the plant crop row (about 5 cm) at very high ground speeds (up to 11 km/h) and chisel or subsoil a field very close to buried drip irrigation tapes without damaging them (Abidine et al., 2004). Pérez-Ruiz et al. (2010) reported that precision transplanter and drill seeder positioning (<0.04 m below) are possible only with RTK-GPS auto-guidance based systems and observed that the use of RTK-GPS auto-steering can result in significant cost savings for vegetable producers.
Several researchers have been working to develop automatic systems to detect separated (i.e., non-occluded) plants from the background scene and to determine different weed species for optimizing and simplifying agricultural work, or for creating weed maps (Pedersen, 2001, Søgaard, 2005, Persson and Åstrand, 2008, Christensen et al., 2009, Staab et al., 2009, López-Granados, 2011). Machine vision-based guidance and weed detection systems have been developed mainly to make more effective use of pesticides, either for band spraying along a crop row or detecting individual weed or crop plants for treatment (Thompson et al., 1991, Marchant et al., 1997, Kouwenhoven, 1997, Tian et al., 1997, Miller and Paice, 1998, Tillett et al., 1998, Fennimore et al., 2010).
A geospatial crop seed or plant map may be a good alternative to real-time weed sensing for use in removing intra-row weeds (Griepentrog et al., 2003, Griepentrog et al., 2005). Ehsani et al. (2004) retrofitted a four-row vacuum planter with a centimeter-accuracy RTK-GPS system and mapped corn seeds during planting. The seed map coordinates obtained were within an average distance of 3.4 cm of the crop plants after germination. Sun et al. (2010) developed and evaluated a centimeter-level accuracy transplant mapping system for precision geospatial mapping of vegetable crops. Their row-crop transplanter, modified for RTK-GPS mapping, had a mapping accuracy of 2 cm on average, with 95% of crop plants mapped within a distance of 5.1 cm from their true location. Recently, a system for geospatial mapping of crop plants using a RTK-GPS automatic guidance system mounted on the tractor without a second GPS system on the planter has been developed, greatly reducing the total equipment cost of the system with only a ∼1 cm accuracy penalty (Pérez-Ruiz et al., 2011 reported a mean accuracy of 3.2 cm compared to the 2 cm accuracy reported by Sun et al., 2010).
Many investigations have attempted to develop mechanical intra-row weeding systems with varied levels of success. These methods can be characterized by the level of technology utilized. Low-tech implements are based on some type of physical property differences between crop and weed that can be exploited for weed control, similar in concept to a selective herbicide or the use of flame weeding in cotton, but based upon mechanical methods. The success of their performance is highly dependent on the crop vs. weed selectivity of the physical factor exploited (Van der Weide et al., 2008). For example, finger weeders place small metal tines into the soil at a shallow depth close to the crop plant and rely on the crop plant having a stronger attachment to the soil than the weeds due to a greater crop planting depth or a larger root system because the crop plants are older than the weeds. For direct seeded crops, the dependence on the weeds being younger or weaker than the crop plants causes the performance of these systems to be erratic in commercial production systems. High-tech tools are equipped with electronic systems for row centering or weed detection. For instance: Nørremark et al. (2008) developed a complex weed control system that required three GPS mission planning files, one each for the tractor, the side-shifting cultivation sled, and the third for the rotating cycloid hoe mechanism. The cycloid hoe consisted of a complex linkage designed to guide eight, sigmoidally shaped soil tillage tines along a non-linear cycloid looping pattern in the soil designed so that the looped path drove the soil tines into the intra-row zone to kill weeds between crop plants. Additional mechanical latching linkages were used to alter the cycloid path of individual tines in order to avoid killing crop plants along the row. The system was evaluated in a simulated crop row using 102 plastic sticks as artificial crop plants hand-placed 0.2 m apart along a 21 m single row where a ruler was used to determine the stick placement between the statically determined geo-referenced endpoints. The mean distance and 95% confidence interval about the mean between the tine and the plastic sticks ranged from 47 mm ± 37 mm to 80 mm ± 42 mm in simulated planting and weeding trials. Tillett et al. (2008) constructed an experimental implement based on a vision-guided inter-row steerage hoe system with two rotary disc cultivators. The rotating hoe blade system had a fixed section cut out from each disc to avoid crop damage. The computer vision system was used to locate the crop plants along the row and attempted to match the phase of the rotating hoe to the crop spacing in order to kill intra-row weeds and spare crop plants. Like the complex control system designed by Nørremark et al. (2008), this design also required a complex feedback control system for accurate hoe positioning.
Some work has been done on autonomous vehicles with real-time robotic weed control systems that navigate through the field, detect, and remove any weeds found (Åstrand and Baerveldt, 2002, Jørgensen et al., 2007, Nørremark et al., 2008, Van Evert et al., 2011). These researchers believe that challenges for robotic weed control are related to: (i) the diversity of the commercial agricultural environment, where differences in weed species and abundance leads to erratic performance in weed detection, (ii) automation and control technology that must respond to changes in terrain levels or static and dynamic obstacles, and (iii) safety in the interaction with both the environment and field workers, e.g. autonomous systems must know when to stop in an emergency.
The aim of this work was to investigate the performance of an automated, intra-row weed knife path control system, where the real-time control input was based on an RTK-GPS geoposition system, an odometry sensor, and an automatically generated GPS map containing the individual geoposition of the crop plants. The specific objectives were (i) to develop a mechanical intra-row weed knife system suitable for RTK-GPS control based upon a crop plant map, (ii) to develop a real-time control system designed to precisely control the path of weed knives operating in the intra-row zone so that they automatically circumvent the crop plants without damage, and (iii) assess the performance of the automated weed knife path control system under standard field conditions in California.
Section snippets
Intra-row knife weeder design
An automatic intra-row weeding machine was designed using a pair of intra-row mechanical weed knives similar in concept to the thinning knife used in the commercial vegetable crop thinner developed by Eversman (Kepner et al., 1978) but modified for precision intra-row weed control and RTK-GPS actuation, Fig. 2. Each intra-row weed knife blade (shown in red in Fig. 2, Fig. 3) was constructed from a 6.4 mm thick plate of hardened tool steel (model Aristocrat D-2, air hardened to Rockwell 60,
Results and discussion
An automatic intra-row weed knife control system, which utilized a GPS crop plant map and a RTK-GPS -based real-time control system to determine the geospatial position of the weed knife blades with respect to each mapped tomato plant in the field, was successfully developed and operated in a processing tomato field. The system successfully controlled the path of a pair of weed knives in the intra-row zone B in the center of the row and minimized intrusion into the close-to-crop zone C where
Conclusion
An automatic, intra-row, weed knife control system, which utilized an automatically generated GPS crop plant map to determine the geospatial position of each tomato plant and on-board RTK-GPS for monitoring the mobile system’s current geoposition in real-time, was successfully developed and operated in a processing tomato field in California. The system was specifically designed to automatically control the path of a pair of mechanical weed knife blades in real-time as they travelled along the
Acknowledgements
The research was supported in part by the Specialty Crop Block Grant program of the California Department of Food and Agriculture. The authors thank Burt Vannucci, Loan-anh Nguyen, Garry Pearson, Jim Jackson, and Mir Shafii of UC Davis, and Claes Jansson and Tord Holmqvist at SWEMEC in Woodland, CA for technical assistance.
References (51)
- et al.
Agricultural robotic platform with four wheel steering for weed detection
Biosyst. Eng.
(2004) - et al.
Robotic weed control using machine vision
Biosyst. Eng.
(2002) Intra-row mechanical weed control – Possibilities and problems
Soil Tillage Res.
(1997)Precise tillage systems for enhanced non-chemical weed management
Soil Tillage Res.
(2007)- et al.
Row-following accuracy of an autonomous vision-guided agricultural vehicle
Comput. Electron. Agric.
(1997) - et al.
Instrumentation and method for high accuracy georeferencing of sugar beet plants
Comput. Electron. Agric.
(2007) - et al.
The development and assessment of the accuracy of an autonomous GPS-based system for intra-row mechanical weed control in row crops
Biosyst. Eng.
(2008) - et al.
Classification of crops and weeds extracted by active shape models
Biosyst. Eng.
(2008) Weed classification by active shape models
Biosyst. Eng.
(2005)- et al.
RTK-GPS mapping of transplanted row crops
Comput. Electron. Agric.
(2010)
Potential for automatic weed detection and selective herbicide application
Crop Prod.
A robotic system for plant-scale husbandry
J. Agric. Eng. Res.
Inter-row vision guidance for mechanical weed control in sugar beet
Comput. Electron. Agric.
Mechanical within-row weed control for transplanted crops using computer vision
Biosyst. Eng.
A review of spectral properties of plants and their potential use for crop/weed discrimination in row-crops
Crop Prot.
Autoguidance system operated at high speed causes almost no tomato damage
Calif. Agric.
Development and field evaluation of a field-ready soil compaction profile sensor for real-time applications
Appl. Eng. Agric.
An agricultural mobile robot with Vision-Based Perception for mechanical weed control
Autonomous Robots
Manual of Geospatial Science and Technology. 2nd Ed.
Site-specific weed control technologies
Weed Res.
Seed location mapping using RTK-GPS
Trans. ASAE
Evaluation and economics of a machine-vision guided cultivation program in broccoli and lettuce
Weed Technol.
Cited by (94)
A precise crop row detection algorithm in complex farmland for unmanned agricultural machines
2023, Biosystems EngineeringIntelligent intra-row robotic weeding system combining deep learning technology with a targeted weeding mode
2022, Biosystems EngineeringCitation Excerpt :The other type is the swing-type weeding device of Steketee and Frank Poulsen, which was proposed by Perez-Ruiz. Experiments in tomato fields demonstrated that the weeding device successfully avoids crops at speeds of 0.8 km h−1 and 1.6 km h−1 (Manuel Perez-Ruiz et al., 2014; Perez-Ruiz et al., 2012). The above equipment involves the exploration of control technology and intelligent technology in intelligent weeding equipment.
An identification algorithm of lateral correction amount for the weeding components in paddy fields based on multi-sensor fusion
2024, Measurement Science and TechnologyOn the Importance of Precise Positioning in Robotised Agriculture
2024, Remote Sensing