ABSTRACT 7th International Symposium on Environmental Concerns in Rights-of-Way Management
TITLE:	Reducing Herbicide Use with Machine Vision Technology
AUTHOR(S):	Harvey A. Holt - hholt@fnr.purdue.edu Department of Forestry and Natural Resources Purdue University W. Lafayette, IN, USA Richard D. Dirks Department of Forestry and Natural Resources Purdue University W. Lafayette, IN, USA Rita L. McKenzie Department of Forestry and Natural Resources Purdue University W. Lafayette, IN, USA Thomas N. Jordan Department of Forestry and Natural Resources Purdue University W. Lafayette, IN, USA Fred Whitford Department of Forestry and Natural Resources Purdue University W. Lafayette, IN, USA
Operator controlled spot spraying is an alternative to broadcast application that reduces the amount of herbicide sprayed. But this method suffers from the limitations of operator capabilities and fatigue. The Patchen WeedSeeker uses light reflectance to detect and automatically treat vegetation. By using individual independent spray units equipped with sensors to detect vegetation, a significant reduction in herbicide use can be realized. A set of three plots of varying weed densities of Russian thistle (Salsola iberica) was treated to determine the effectiveness of the sensitivity controls. These tests were conducted using spray units calibrated to apply 296 l/ha (31.7 gpa) as a broadcast application. WeedSeeker application ranged from 51-145 l/ha (5.5 and 15.5 gpa), depending upon the sensor’s sensitivity setting and weed density; a reduction of 50-77% over broadcast application. Tests were conducted to compare the WeedSeeker with operator-controlled application. These plots were located on the KB&S Railroad near Lafayette, IN and included a wide variety of weed species and densities. Sprayers were positioned 30.5 cm (12 in) apart and calibrated to spray 224 l/ha (24 gpa). Rates for the WeedSeeker ranged from 34 to 180 l/ha (3.6 to 19.3 gpa) while operator-controlled application rates ranged from 173-224 l/ha (18.5 to 24.0 gpa); an average savings of 61%. Percent bare ground for the WeedSeeker treatments was not significantly different from results with the operator controlled application. Tests were made to determine the effect of plant size on detection. A grid of leaf circles of four diameters [3.9, 6.4, 9.6 and 13.4 cm^2 (0.6, 1, 1.5 and 2.1 in^2)] was randomly placed along the rails with the sprayer passing over the circles. Three replicates were conducted using three different speeds, 19.3, 24.2 and 29 kph (12, 15 and 18 mph). The smallest leaf circles were treated 52% of the time when averaged over the three speeds. Average percentages for the other circles sizes ranged from 83 to 97%. Ref#: Abs-1-157