A MICROPROCESSOR CONTROLLED TECHNOLOGY

A MICROPROCESSOR CONTROLLED TECHNOLOGY

TO SELECTIVELY SPOT SPRAY WEEDS

W.L Felton¹, A.F. Doss², P.G. Nash³, and K.R. McCloy⁴

Erosion and pesticide use are environmental issues of increasing community concern and cannot be ignored. Although herbicides have assisted the economic viability of farmers and helped reduce the risk of soil erosion, there are many situations where weed control with herbicide is more expensive than cultivation and their use may be seen as a potential ecological hazard.

Cultivation is still the primary method of fallow weed control in most grain production regions. However, in the last decade there has been a substantial increase in herbicide use, especially immediately before sowing, and strategic spraying during extended wet periods when traditional farming methods are less effective.

Weed control is important as even sparse weed population can decrease the moisture and nutrients available to subsequent crops and reduce productivity. In controlling scattered weeds, most of the herbicide sprayed from a standard spray boom will be deposited on the soil or litter. The potential to reduce both herbicide costs and the possibility of environmental problems by selectively spraying only the weeds is therefore considerable.

Devices designed to discriminate green vegetation from other ground covers by use of physical contact such as the "weed wipers", rely on the green vegetation being in distinct physical planes. In practice this rarely occurs because weeds, particularly those with a prostrate growth habit, may remain within that stubble canopy. Therefore, another method of discriminating green vegetation from other surfaces was required and an optical method was considered the most promising.

THEORETICAL BASIS OF THE TECHNOLOGY

Solar radiation incident on a surface is partially reflected from, transmitted through, or absorbed by the surface. Most incident solar energy is transmitted into green vegetation. It is then selectively absorbed in the blue and red wavebands by the chlomphyus and strongly reflected in the near-infrared (near-IR) waveband by the complex internal structure of the plant (Swain and Davis 1978).

Green vegetation absorbs strongly in the red waveband and strong reflection in the near-IR waveband compared to both dead vegetation or soil (Fig. 1). These two wavebands are often used to achieve the best discrimination between green vegetation and other ground covers (Tucker 1980).

Fig. 1

Hagger et al. (1984) and Mayhew et al. (1984) developed , portable instruments to estimate green plant density or leaf area index. These measure the radiant flux reflected from a surface (radiance) in the red and near-IR wavebands and, from the measured values, derive a near-IR to red ratio.

This ratio has been found by Mayhew et al. (1984) to highly correlated with green biomass under stable conditions of incident radiant flux (irradiance). Hagger et al. (1984) reported similar results.

The density of radiance depends upon that of irradiance reflective properties of the surface. The density of radiance depends an latitude, season, time of day, atmospheric attenuation, cloudiness and shadow, and the portion of all spectrum being considered. Measurements of radiance for a particular surface may therefore

^{1Senior Research Agronomist, ² Programmer/Analyst, ³ Field Officer, NSW Agriculture, Agricultural Research Center, Tamworth 2340, Australia.
^{4Remote sensing consultant, 2 James Street, Petersham 2049, Australia.

vary considerably. To obtain consistency in these measurements, the level of irradiance must be considered.
If the radiance in a waveband is expressed as a percentage of its irradiance, a property known as reflectance, which is consistent under changing conditions, may be derived. The ratio of near-IR reflectance to red reflectance will similarly be consistent for a given surface under varying conditions.

The objective was to develop and evaluate a reflectance based device that could automatically activate each nozzle of a boom so as to spray only green vegetation. A "weed detector" should sense a field-of-view consistent with the spray width of each herbicide nozzle, and function under a wide range of environmental conditions. It needs to discriminate green vegetation from other surfaces, which may be extremely variable in terms of soil types and conditions, in an automatic and consistently accurate manner.

The red and near-IR wavebands providing optimal discrimination (630-670 nm and 830-870nm respectively) were selected after extensive high resolution spectral analyses of green and dead vegetation, soil, and litter, using a Collins IT IS spectroradiometer. Some types of litter may have responses in the red and near-IR bands that result in the near-IR/red ratio being similar to that of green vegetation. It may, therefore, be necessary to modify the simple ratio discriminant function currently used.

DESCRIPTION OF THE PROTOTYPE

The "weed detector" prototype is shown schematically in Fig. 2. The optical fibre carries reflected energy to the detectors via the red and near-IR bandpass filters. The diffusing plates integrate hemispherical irradiance for transmission through corresponding filters to light detecting diodes. The diode signals are amplified before transmission to an analog-to-digital (A/D) converter, which transmits digital data values to the central processing unit (CPU) for storage, analysis and display. The result of the analysis is a logical decision as to whether or not green vegetation has been sensed. The CPU then activates a solenoid operated nozzle if required.

Fig. 2 Schematic diagram of the "weed detector" prototype

THE COMMERCIAL PROTOTYPE

Using the information gained from evaluating the prototype, a 37 nozzle, 18 metre spray boom was designed and fitted to a four wheel drive chassis. One irradiance sensor pair is mounted above the vehicle cabin. Each independently operated nozzle has a radiance sensor pair and solenoid control valve. The radiance sensor pairs, which have a field-of-view matching the spray pattern, are located ahead of the nozzles and point downwards, negating the need for optical fibres. When the sensors detect a green plant, or an area of green plants, a microprocessor turns the nozzle on for a predetermined period. A central control microprocessor located in the spray vehicle communicates data to each detector and may be used to alter the sensitivity of the system. Detection and activation are virtually simultaneous so that the period from detection to spraying the target is primarily determined by the mean spray velocity. The system may be used as an overall boom spray by selecting a continuous spray option instead of the detection mode.

FEILD EVALUATION

Since March 1990 the commercial prototype has been evaluated in a wide range of conditions in northern New South Wales and southern Queensland, Australia. It was operated at a ground speed of 12 kph and a spray line pressure of 250 kPa, as is common commercial practice. There was a diversity of annual and perennial weeds that exhibit a broad range of growth habits which included Johnson grass (Sorghum halepense), couch (Cynodon dacrylon), Noogoora burr (Xanthium occidentale), Bathurst burr (Xanthium spinosum), mintweed (Saivia reflexa), camel melon (Cucumis myriocarpus), devil's claw (lbicella lutea), thorn apple (Datura spp), milk thistle (Sonchus oleracae), black thistle (Circium vulgare), rattlepod (Crotalaria dissitilflora), wild oats (Avena spp), volunteer sorghum (Sorghum vulgare), liverseed grass (Urochloa panicoides), barnyard grass (Echinochloa spp) and nutgrass (Cyperus spp).

A total of 2681 hectares was treated on 33 farms. The mean saving in area sprayed, relative to an overall treatment, was 90%. On 20 of the farms, more than 95% of weeds were killed and on a further 5, more than 90% were killed.

In the 8 fields where control was less effective this was attributed to:

Weeds too small to be reliably detected.
An inadequate herbicide rate.
Specific weeds not completely controlled by the herbicide mixture.
Rough conditions causing spray boom instability.
Rain after application reducing herbicide performance.

Some of the perennial weed species, for example couch and Johnson grass which were present on 10 sites, can require a high rate of herbicide to control, and are inadequately controlled by cultivation. Excellent control of these weeds was achieved on 9 of these by selectively spraying with the recommended application rate and because a minimal amount was wasted, was cost effective (Table 1). On the 23 sites where annual weeds were the main
problem, and lower herbicide concentrations could be used effectively, there were also considerable cost savings, as shown in Table 1.

Without selective spraying, farmers would have made the decision based on cost, to cultivate25 of the 33 fields. The cost of controlling larger weeds by spraying and the risk of poor control is the main factor influencing farmers to cultivate.

The most important considerations in spraying weeds are usually which herbicide(s) to use, the rate of application, and the method and timing of application. Herbicide selection with spot spraying is as important as it is with overall spraying, however the rates are less crucial. If only 5-10% of a paddock is to be sprayed, there is a negligible difference in cost, for example, between using 1.5 and 2.0 L/ha of glyphosate. Large or chemical tolerant weeds should no longer escape control because of insufficient herbicide. Mixtures should become more prominent, and products which were previously prohibitive in cost on an overall spraying basis, may now be considered.

Table 1. Costs of controlling perennial and annual weeds in fallow by selective or overall spraying, or cultivation}}

		Weed type
		Perennial	Annual
	Number of farms	10	23
	Total hectares sprayed	1004	1677
		(AU$)	(AU$)
Selective spraying ^a
	Range	1.50 - 17.70	0.45 - 10.20
	Average	6.27	3.36

Overall spraying^a
	Range	26.00 - 91.00	8.00 - 36.00
	Average	60.10	20.31

Cultivation
	Average	15.00	15.00

^{a Excludes application costs, approximately $3.00/ha

CONCLUSIONS

The weed detecting spraying system has the potential to increase the adoption of minimum tillage practices by farmers.

Benefits of this technology include the reduction in the volume of spray and consequently the cost of herbicide, time savings because of fewer stops to refill, lighter spray rigs and hence reduced soil compaction, better control of tolerant weeds, and less non-target spraying which reduces potential environmental risks.

Agriculture has been the focus of our effort but there are many situations where weed control is of concern and selective spraying would be beneficial. Other uses for a weed/green vegetation detection system include the control of weeds on railway tracks, road verges, airports and other industrial areas, band spraying between row crops, and the application of water nutrients and other pesticides in intensive agriculture and horticultural crops.

NSW Agriculture have applied for Patents covering this technology in Australia, New Zealand, Canada, United States and Europe. New England Technology Pty Ltd, Armidale, NSW, has been assigned the commercial rights to the technology. The design is being modified to improve performance, production suitability and ease of installation. Prototypes for test marketing have been designed.

REFERENCES

(1) Hagger, R.J., CJ. Stent, and J. Rose. 1984. Measuring spectral differences in vegetation canopies by a reflectance ratio meter. Weed Res. 24:59-65.}