D. K. Giles

Precision weed control system for cotton

A real–time robotic weed control system was developed and tested in commercial cotton fields. The precision weed control system was capable of distinguishing grass–like weeds from cotton plants and applying a chemical spray only to targeted weeds while traveling at a continuous speed of 0.45 m/s. The robot consisted of a real–time machine vision system, a controlled illumination chamber, and a precision chemical applicator. In commercial cotton fields, the system correctly sprayed 88.8% of the weeds while correctly identifying and not spraying 78.7% of the cotton plants while traveling at 0.45 m/s.

PRECISION BAND SPRAYING WITH MACHINE-VISION GUIDANCE AND ADJUSTABLE YAW NOZZLES

A system was developed for precision, ground-based application of foliar sprays to rows of small plants. The system consisted of a machine-vision guidance system that positioned spray nozzles directly above each row of small plants. An electrically actuated mechanical linkage rotated flat-fan spray nozzles about their central axes to vary the yaw angle of the fans relative to the direction of travel. The nozzle rotation effectively changed the width of the spray pattern relative to the direction of travel and allowed the spray band to be adjusted to match the width of the target plants. The system was evaluated on young stands of tomatoes and lettuce by measuring spray deposition on the plants and surrounding soil. The precision system allowed spray application rates to be reduced by 66 to 80% and increased spray deposition efficiency on the target plants by 2.5 to 3.7 times greater than conventional broadcast spraying. Nontarget deposition on soil surfaces was reduced by 72% to 90%. Airborne displacement of spray liquid from the spray boom, as measured by collection strings surrounding the boom structure, was reduced by 62 to 93%. The reduction in displacement suggested that spray drift from precision applications could be significantly less than that from conventional applications.

HERBICIDE MICRO-DOSING FOR WEED CONTROL IN FIELD-GROWN PROCESSING TOMATOES

Weeds within the seedline area and in close proximity to crop seedlings are highly competitive and reduce crop yield. When chemically selective herbicides are unavailable or ineffective, spatially selective removal is required and is often a manual operation. Automation of selective weed removal requires sensing and actuation systems. Image and spectral analysis can discriminate weeds from crop plants while precise dosing methods are required for automated treatment of small, distinct weeds. A pulsed-jet, micro-dosing actuator was developed to apply liquid herbicide treatments. The biological performance of micro-dosing was evaluated in field-grown processing tomatoes with a nonselective herbicide (glyphosate) in mixtures including surfactants and splash-inhibiting polymers. Liquid dose rates were 37 .L/cm2 for treatment areas of 6.3 . 12.5 mm. Pulse durations were 6 to 10 ms. A formulation including a surfactant and a polymer provided efficacy while reducing splash, i.e., “micro-drift” that caused phytotoxicity to adjacent crop plants. When treated with the micro-dosing system, splash-related phytotoxicity produced greater yield suppression than competition from escaped weeds. The results established that use of non-selective herbicides for micro-scale dosing of weeds during early crop growth is a feasible alternative to broadcast application of chemically-selective herbicides. The optimal point of crop and weed growth for using a weed-sensing, one-pass post-emergent control system requires understanding of crop development, weed competition, and the ability of the crop to recover from inadvertent splash of micro-dosed materials.

Finite element analysis of particle and liquid flow through an ultraviolet reactor

A computational fluid dynamics program was used to simulate liquid flow through an ultraviolet (UV) reactor. Salt tracer tests were performed at different flow rates to validate results from the program. Temporal and spatial descriptions of particle trajectories from the flow model were used to quantify flow pathways through the UV reactor. Particle trajectory information was then used in a separate numerical program to estimate temporal distribution of particles exiting the domain during discrete time steps for the flow solution. From these data, the exit time for particles transported through the modeled flow domain were compared with experimental tracer results. At low flow rates, the simulation results were within 95% confidence intervals established from the experimental data. At higher flow rates, the simulation results tended to predict a shorter exit time for particles than that observed experimentally with salt tracer studies. The results of this research can be used to compare UV disinfection efficacy on low transmission liquids without the need for extensive experiments or construction of prototypes.

DISCRIMINATING WEEDS FROM PROCESSING TOMATO PLANTS USING VISIBLE AND NEAR-INFRARED SPECTROSCOPY

Nightshade weeds in California processing tomato fields were studied to determine if visible and near-infrared (400 to 2500 nm) reflectance spectroscopy could be used to discriminate nightshade weeds from tomato plants. Optical absorbance data in the 2120 to 2320 nm region gave the best classification rate between weeds and tomato (100%) using narrowband hyperspectral models and canonical discriminant analysis. Narrowband hyperspectral models in the visible region also gave good classification performance (95% accuracy); however, broadband color-based models were only 75% accurate. The effect of reduced analog to digital conversion resolution on classifier performance was also studied.

Digital device and technique for sensing distribution of spray deposition

Assessment of spray deposition often relies on the use of water- and oil-sensitive paper or other stain-based techniques to document extent of coverage and distribution. The procedure requires the placement, collection, scanning, and post-processing of deposition stains on cards, and is time consuming and labor intensive. An electronic sensor, coupled with a wireless network for transmission of in situ data, would have great advantages over current methods typically used in spray deposition research. A prototype technique and sensor for detecting and quantifying spray deposition were developed and evaluated. The device is capable of sensing the presence and location of multiple droplets on a sensor surface, is reusable and potentially inexpensive. The sensor detects the presence of discrete fluid droplets on the sensor surface when droplets complete electrical circuits positioned in an array. The sensitivity of the system is controlled by setting a reference (threshold) voltage in a voltage comparator circuit, while spatial sensitivity is determined by the physical separation of sensing points on the sensing array. The sensor differs from previously reported designs in that the deposition is quantified as a series of discrete, digital indications instead of a spatially integrated response. This capability allows spatial distribution of spray, not just an estimate of quantity, to be determined. The sensor was tested using spray applications of fine, medium, and coarse ASAE standard sprays. The performance of the sensor prototype, defined as measurement of covered area of stain deposition, was best when detecting coarse sprays. Future development will include fabrication of a sensing array with a higher spatial resolution of sensor pads, incorporation of more elegant algorithms for detailed scans of the sensing surface, and interfacing the sensor with a wireless communication system.

Direct nozzle injection of pesticide concentrate into continuous flow for intermittent spray applications

A direct nozzle injection system was developed to intermittently inject concentrated solutions into continuous carrier liquid flow through a straight-stream spray nozzle used for targeted roadside spraying of post-emergent herbicide during pre-emergent herbicide application. The injection system was based on a 12 VDC direct-acting electrical solenoid valve with a 0.56 mm valve orifice and metering plate with a 0.2 mm diameter orifice. A conductivity-based sensor was used to measure the instantaneous concentration of NaCl tracer simulating a pesticide solution. Injection pulse durations ranged from 10 to 100 ms into carrier flows of 1.5 and 2.6 L/min from 1.55 and 2.18 mm nozzle orifice diameters, respectively. Lag times between initiation of the injection valve actuation and emission of the concentrate material from the spray nozzle were on the order of 25 ms. Concentrations of 1% (v/v) from injected solution into the flow emitted from the nozzle could be achieved within 100 ms after valve actuation. Increasing the injection pulse duration did not reduce lag time nor increase the temporal rate of concentration increase in the emitted spray; however, increasing the injection pressure increased the rate of concentration increase. Analysis of the injection event, using standard mixing criteria, determined that the injection mixing events were Gaussian in nature and did not represent ideal plug flow or short-circuiting events. For an intermittent, target-detecting system with a detector-to-nozzle distance of 4 m and a ground speed of 5 m/s, the direct nozzle injection system is a feasible configuration for spot spraying if the sum of detection time and time of flight for the emitted spray is less than 800 ms. For a prototype machine vision-based roadside sprayer, detection and spray flight times were less than 67 and 400 ms, respectively; therefore, feasibility of spot spraying using at-nozzle injection was established.

INJECTION MIXING SYSTEM FOR BOOMLESS, TARGET–ACTIVATED HERBICIDE SPRAYING

An injection mixing system for herbicide solution was developed for a target–activated spray system designed for roadside weed control. Commercial components were used for fluid handling, and two interface techniques were developed for communication between the target sensing sprayer and the injection system. An electronic technique used digital scaling of the relative weed load to characterize and initiate changes in relative injection rate of active ingredient. The alternative mechanical technique relied on fluid coupling between the spray emission and the injection systems to maintain the desired concentration of active ingredient in the spray mix. Performance of the injection system was evaluated using salt tracer as the injection solution and in–line liquid conductivity measurements to determine concentration. Response, stability, and steady–state accuracy of the system were quantified over a 12:1 turndown ratio at command rates of 0.5 to 10 Hz. Both interface techniques maintained desired tracer concentration within 10% of the target level when subjected to highly variable weed loads. The electronic technique provided faster response and the possibility for feedforward control. Liquid pressure at the nozzle manifold was maintained during the rapid changes in spray emission by using 3–way valves to shuttle flow between the spray nozzle and a bypass line. Nozzle pressure remained within 10% of the desired value during the dynamic testing. The fluid handling and injection control systems were transparent to the operator and required no changes to the operator interface typically used for conventional spraying.