from Agrii Company News
A highly accurate, fully-automatic, ground-based weed mapping system enabling growers to target key aspects of their black-grass management with the greatest precision should be available for commercial use within the next two years, according to the latest report from pioneering eyeWeed project co-funded by Innovate UK (formerly the Technology Strategy Board). What is more, it is set to form the basis of a season-long platform for monitoring and mapping a growing range of valuable precision farming parameters.
The project consortium of commercial and academic partners led by Agrii and including the University of Reading, Concurrent Solutions LLC, Knight Farm Machinery, Syngenta, and Patchwork Technology has been successfully turning the University’s original proof of concept into a practical system for reliable farm black-grass mapping use over the past five years.
The advanced eyeWeed prototype being put through its paces this season comprises six spray boom-mounted cameras linked to sophisticated computer software that can accurately map black-grass patches within wheat crops in mid-June at much higher resolution than is possible with current aerial imagery. Importantly too, it can produce black-grass infestation maps in real-time without any extra operations and with none of weather-related limitations of drone-based or satellite detection.
“We have developed the system in detailed work on a number of fields across several farms over three full seasons to map black-grass patches with great accuracy in wheat during T3 spraying,” explained Agrii decision support services manager, John Lord who leads the project. “Unlike other approaches, no field walking is required to ‘ground truth’ the weeds identified and all the work can be done as part of normal field operations, whenever the weather is suitable for spraying.
“The maps we are producing can be automatically employed in management zoning as part of precision farming systems for variable seed rates as well as targeted pre-, peri- or post-em patch spraying. “In all but the worst cases, black-grass is concentrated into well-defined field patches. So it makes sense to concentrate higher seed rates and the most robust autumn herbicide stacks and sequences on these wherever possible, rather than employing them across the whole area. Not least because a patch-spraying programme can increase your gross margin by £50/ha or more as well as delivering valuable environmental benefits at a time when both are becoming increasingly important to most growers.
“The maps can also be used for automatic summer patch-spraying with glyphosate, if necessary, and for greater precision in autumn cultivations and stale seedbed management where feasible.” John Lord added. “At the same time, they provide the best possible field-by-field records for planning future black-grass management and tracking the success of control strategies, not to mention the development of resistance.”
For the greatest convenience and value, John Lord explains that the eyeWeed data will be made available to growers and their agronomists through the Agrii Precision Services (APS) portal in the most effective integration with the company’s SoilQuest soil mapping, MetQuest weather station data and other precision agronomy and decision support services.
While black-grass management has been its initial driver, he stresses that the system emerging from the project offers huge potential for use in a range of other mapping and crop management applications linked to sprayer work throughout the year.
Currently being explored by the team are the summer use of the camera system to map black-grass in barley and development of alternative algorithms to identify other weeds such as ryegrass, wild oats and brome in wheat.“We’ve also designed the unit to take another sensor at each camera housing, so it could be employed to provide NVDI maps for variable nitrogen application or any of a number of other sensing technologies under development too,” he pointed out.
“Essentially, we’re producing a ground-based multiple-sensing platform that can be added to modern sprayers to enable the most accurate crop monitoring and mapping throughout the season as part of their normal operations. “Within the next two years we’re confident of being able to introduce the most robust, accurate and work-proof black-grass mapping service at a price that will cover its initial costs in as little as three seasons.
“As we develop and validate the necessary additional sensing and analytical software, we see eyeWeed being employed as a regular tool at no extra effort and for little extra cost throughout the spraying season to provide invaluable, near-to-real-time crop information for the most effective precision agronomy.”
We are fabricating an autonomous robotic platform for the University of Reading in the United Kingdom. This system will be used in field research to be delivered early 2019. This platform will utilize our machine vision system to identify weeds in horticultural crops. As part of this system we will provide a leaf-specific herbicide applicator capable of applying systemic herbicide exclusively to the weeds in the presence of the vegetable crops.
In support of this effort we have provided to their research a number of custom laboratory instruments including a leaf-specific herbicide application testbed. This system is capable of delivering precisely measured doses of liquid herbicide with sub-microliter precision. In addition to precise dosages, this system can move the applicator at precisely controlled speeds using a linear actuator and a custom software application. The images below show the effects of leaf specific herbicide applications on various growth stages of cabbages.
Another customized instrument provided as part of the testbed was a cellular linear wind generator to produce straight-line winds over a range of wind-speeds to simulate the types of ground-effect winds experienced in the field. This was used to produce calibration tables for adjusting herbicide targeting under a variety of field conditions
The robot being delivered to University of Reading is a prototype platform based on a design we have used for our own field research. It will be equipped with an onboard row navigation system developed and demonstrated as part of the previous USDA Phase I SBIR grant. This platform is equipped with four-wheel drive and four-wheel steering. It is powered by a gasoline generator over electric battery power system for long unattended field operations. It can carry the geo-referenced image collection and storage processor built for the eyeWeed Mapping program, a herbicide applicator module, and a variety of crop scouting and monitoring sensors.
Yet another object of the present invention is to provide a two-part method of calibrating incoherent fiber optic bundles to transmit visual or infrared coherent imageS which consist of the following: 1) provide a fast and simple method for obtaining the coordinates of the fibers on the input and output face and how they relate to each other based on an initial camera and optical interface and 2) provide a method to allow the output fiber coordinates to be transformed for alternative cameras, optical interfaces, or the refocusing that is required when transmitting infrared images. Part 1 would be performed by the fiber bundle Supplier using a simple hardware apparatus and Software algorithms defined herein. Part 2 would be performed by the end-user using software algorithms along with the initial fiber coordinates and images as explained herein.
In the first years of their existence, Bevilacqua Research Corporation (BRC) in Huntsville AL contracted the help of Concurrent Solutions in the pursuit of new business. One of their first contract awards was a Phase I SBIR from the Army Research Labs (ARL) to develop an automated decision aids system. This system was to support Battle Damage Assessment of tank columns in battle, using visual cues to determine the order of battle. Concurrent Solutions was charged with the development of the software applications for this effort.
A number of different AI tools were
considered including Cluster Analysis, Adaptive Resonance
Theory, Neural Networks, Expert Systems, and Conceptual Graphs
(John Sowa). It became clear that different elements of this
problem needed different AI approaches so we developed a
software tool for combining these various types of decision
aids into a single decision aid system. In addition to the AI
tools, standard phyics-based modeling was integrated to reach
realistic, "common sense" conclusions when there was
sufficient data available. This was the first AI tool
combining multiple intelligent agents.
Dialectic Neural Network (DNN) -.
One of the first applications of the IMADS was in the
development of a DNN for the Army Research Laboratory Contract
No. DAAL01-94-C-0040 with Bevilacqua Research Corporation in
Huntsville, AL. This application supports Battle Damage
Assessment against ground armored vehicles.
Intelligent Bayesian Network (IBN) - A statistical network using apriori knowledge to deduce the most likely interpretation of partial and questionable information. Used as a decision support system in strategic planning applications.
Multiple Neural Network (MNN) - A network of neural networks (NNs) in which a pure bottom-up solution is sought, but in which the specific features being trained into the network weights must be precisely controlled. The first layer of NNs is trained separately to ensure the capture of the correct feature set. The outputs of these feature detector NNs are then passed to one or more second-layer NNs for higher level processing. This was an implementation of a Deep Neural Network (DNN).
Adaptive Semantic Network (ASN) - Following the methods of John Sowa, this application is a collection of Conceptual Graphs (CGs) connected by relational database operations into an overall semantic network. ASNs are used in various artificial understanding applications.
The ghillie suit is worn by snipers and other military special forces. It is the "swamp monster" suit that blends in with the environment. While there are many varieties of ghillie suits commercially available for hunters, the ghillie suits worn by special forces are customized by their wearers, typically incorporating vegetation and other materials taken from the environment in which they intend to become virtually invisible to the eye and standard surveillance cameras.
In modern warfare, however, the threat of being detected is no longer limited to the visible spectrum. Heat sensing sensors such as FLIRs (Forward Looking InfraRed) or thermal sensors can detect a person's body heat. More recently, millimeter-wave radars are being employed to detect humans, even through walls and foliage. In order to remain invisible, the ghillie suit would need multi-spectral capabilities. This was the task handed to Teledyne Brown Engineering (TBE) by Special Forces (SOCOM) and U.S. Army Natick.
TBE set about fabricating and testing a wide variety of textiles, dyes, coatings, and cut patterns in various combinations to find the right combination that would provide adequate visible, thermal and radar signature suppression. The problem quickly became overwhelming. At its most complex there were 27 different parameters that could be adjusted to affect the emissive and reflective properties of the material across the three spectral bands. Unfortunately changing any one of them tended to affect the properties of the others.
It was at this time that Concurrent Solutions was contacted to help organize and interpret the enormous amount of data being generated. Our solution was to create a more effective way to analyze and compare the material properties and test results. Our solution was a visualization tool that allows the user to view and analyze complex data in a unique way.
The Multivariate Visual Analysis Program (MVA Pro) permits the user to simultaneously view six or more dimensions of a data set of arbitrary dimensionality. This tool reads a standard CSV (comma separated variable) file or Excel spreadsheet, and programmatically creates a GUI interface. From this interface the user can select a subset of the data features to view as well as choose the graphical elements or dimensions in which they are displayed. In addition to displaying the data, the user can invoke a number of analytics tools to sub-select and analyze the properties of particular data clusters of interest.
The graphical elements of the multi-dimensional display include X, Y, and Z (the Cartesian coordinates), color, and the length and diameter of the data "point" we called the bloboid. If the length of the bloboid is greater than its diameter, it resembles a fat cigar. As the diameter parameter begins to grow larger than the length parameter, it transitions through a spheriod shape, coming to look like an M&M candy or "frisbee".
Through keyboard controls, the user can fly through the data set, view and analyze it from any perspective. Subtle trends and patterns not detected by normal statistical analysis can be readily observed in the animated graphical display. Goal states can be dropped into this geometric data space and bo compared to actual measurements to support (in this application) fabrication decisions. Queries can be applied to the database with their results displayed in tabular as well as graphical forms.
MVA Pro has many other capabilities, including an automated normalization function for dealing with non-numeric (categorical) data features. When needed additional dimensions can be integrated into the graphical display by replacing the blobloid with a toriod. This gives the user the capability to match additional features to the major and minor radii of the toriod, three more features can control the orientation of the toriod. Finally, in extreme cases the toriod can be made to spin and rotate to represent even more features. MVA Pro can provide a deeper intuitive understanding of complex data than can be obtained by conventional analytic methods.
SAIC is the lead contractor for the U.S. Missile Defense Agency (MDA), validating and distributing imagery data for seeker and sensor algorithm developers as part of Project Hercules. The mission of this effort is to address and resolve some of the toughest algorithm issues associated with ballistic missile defense.
according to Defense Industry Daily staff
One of the hard problems in missile defense is how to deal with decoys. These are objects dispersed from an inter-continental ballistic missle (ICBM) to confuse and overload the data processing functions of surveillance and tracking sensors of the defense system. By the time most anti-ballistics missiles (ABMs) are launched, a MIRV (Multiple, Independent Re-entry Vehicle) missile will have split into its component warheads and the decoys will have been deployed (as many as 200 for each lethal reentry vehicle (RV)). You might wonder why the offense doesn't just add more RV to their missile. The fact is, modern ICBMS have space for more MIRVs than they carry due to strategic arms limitation treaties, so its easier, and cheaper to put a few decoy MIRVs in the missile than it is to build a new interceptor to counter each MIRV. You could MIRV the kill vehicles, but that’s not yet an option with smaller missiles. The other way to fight this multi-headed hydra, of course, is to get really proficient at figuring out which objects are decoys.
Enter the US Missile Defense Agency’s Project Hercules, a national effort to develop related algorithms and battle management concepts. Robust detection, tracking, and discrimination algorithms useful against targets in all phases of flight; a physics-based decision architecture that applies advanced decision theory to future Ballistic Missile Defense (BMD) System Command, Control; and Battle Management Communications (C2BMC) concepts are all involved. MDA says it “Focuses national expertise on discrimination for the benefit of all BMD System elements,” and the algorithms et. al., will support spiral development via insertion and upgrade of its spinouts in other systems.
Proceedings Volume 1050, Infrared Systems and Components III; (1989); doi: 10.1117/12.951424
Due to the backgrounds and experience of some of our members, Concurrent Solutions was invited to join the SAIC team: specifically over 15 years experience with tracking and target discrimination algorithm development for strategic defense systems. It is important to understand the problem of detection and tracking in the strategic defense problem domain. The objects being tracked by a space-based strategic sensor are not resolved. This means that the size and shape of the image on the sensor focal plane wholly due to the physical characteristics of the optics of the sensor (called blur spots) rather than the size or shape of the target. While this makes target identification more challenging it offers an advantage for solving another issue: detection of closely spaced objects (CSOs).
CSO's are a critical issue for strategic infrared systems. Whenever a sensor's FOV is impinged with a large number of densely spaced objects the blur spots begin to merge. This corrupts both the spectral band measurements and the spatial measurement essential to state (impact) estimation and object type identification (called discrimination in strategic defense). Concurrent Solutions' main task for SAIC was to develop a method to recognize when two or more objects are merged into a single blur spot (technically this occurs when the multiple objects are close enough that their combined blur spots produce a single peak).
The algorithms used in this project are based on being able to detect deviations in the shape of the blur spot from that observed for a single target. This method was demonstrated to achieve CSO recognition for two target separated by as little as 1/3 the width of the typical blur spot. A member of Concurrent Solutions was an author on a paper describing this method as the likely preferred approach. This was verified over a decade later on Project Hercules.
Algorithm for Detecting CSOs using Radial Variance
More than 6,000 American Soldiers have been killed in combat in Iraq and Afghanistan since 2011 and over 50,000 have been wounded. In an effort to offset these staggering losses, the Geotechnical and Structures Laboratory (GSL) of ERDC (Engineering Research Development Corp) continues to develop and test protective technologies with the single goal of saving lives.
As part of a Broad Agency Announcement (BAA) contract, Concurrent Solutions was tasked to devise an alternative to the Hesco barrier (bastions), which is basically a canvas lined wire-mesh box that can be unfolded and filled with whatever materials are locally available. They are used to build protective walls around fixed facilities in forward (or downrange) areas.
An issue with using Hesco barriers is that they require the use of heavy equipment to assemble them. Also, once an enclosure has been constructed it is difficult to take down or modify and the wire-mesh containers are not reusable. Finally when a wall built from Hesco containers is damaged by mortar or rocket fire, it is difficult to repair.
The list of objectives given to CS-LLC for the alternative protective system included, (1) It must erectable with human labor only (i.e. no heavy equipment), (2) the time to erect a wall must be fast, (3) the components must be reusable and damaged walls must be relatively easy to repair. Our solution was the Modular Protective System (MPS).
The MPS is comprised of an expandable frame (like a pop-up sports chair) that holds epoxy fiberglass and flexible concrete panels developed ORIGINALLY by US Gypsum. In addition to the expandable frames we designed leveling pads and specially designed "z-bar" channels to hold the panels without creating stress points. We solicited Edwards Design and Fabrication to fabricate the MPS systems and replacement parts.
The MPS have been extensively tested by ERDC in live-fire demonstrations, held at Fort Polk, LA. They tested these rapidly deployable protective structures for deployed troops, focusing specifically on small Combat Outposts (COP) of 300 personnel or less. Researchers successfully demonstrated that the MPS increases protection without increasing the logistics and necessary equipment needed to deploy the protection.
"This event demonstrated performance in an operational environment," said Nick Boone, a research mechanical engineer in GSL's Survivability Engineering Branch. "Each technology was employed in a manner that might be experienced in a Combat Outpost scenario. The physical COP site really helped solidify everyone's understanding of how these technologies would be used in remote deployed areas.
"In the past, demos only emphasized the experimental performance of new technologies," continued Boone. "It is equally important to demonstrate the extreme conditions where new technologies might be used—it helps everyone understand the true problems that need to be solved."
COPs have limited manpower and logistics resources, and must be rapidly established in hostile environments while remaining safe, secure and potentially mobile. This is an objective that is not being filled by the current primary means of protection, including the use of soil berms, earthen revetments, Hesco or concrete barriers, heavy timber or sandbags. The MPS allows Soldiers to achieve equal, and in many cases superior, protection in mere hours versus the several weeks it can traditionally take to deploy full protection. All components can be easily airlifted to remote locations and set up with minimal effort and manual labor.
By detonating explosive charges in close proximity to each of the technologies during the Fort Polk demonstration, researchers were able to simulate the ability of the protective measures to withstand actual combat conditions. Some of the experiments, such as a simulated vehicle bomb along a perimeter wall, used bare charges to imitate ideal blast loading conditions. Others used threat mortars and rockets, emulating a complex attack on the outpost. To emphasize the focus of the protective technologies, plywood dummies were also installed through the outpost to simulate troops.
"The goal is to save Soldiers' lives," said Boone. "Many of the protective structures were occupied by human simulates to provide a reference for what might happen to anyone within that vicinity during the attack. In addition, several of the simulates were installed so they were fully exposed to weapon fragmentation and blast characteristics. This provides a good basis for comparison when analyzing structure performance." As a result of this effort. members of Concurrent Solutions were included as inventors on an ERDC patent and CS-LLC retained royalty rights for the sale of these systems.
Currently MPS Starter Kits and replacement parts have National Supply Numbers and are available for sale to NATO countries.