Potential of image processing for determining the control rates of common cocklebur (Xanthium strumarium L.) in flaming appications

The objectives of this research are to compare flame applications to Xanthium strumarium L. under controlled and field conditions and to determine whether image processing can be used effectively to determine weed control rates. For this, pot and field trials were established. Flaming was performed at five different doses of LPG at the growth stages of the weeds. Visual control rates were determined at DAT1, 7, and 14, and dry matter and plant height measurements were made. Using log-logistic models, dose-response curves were found. Then, the images taken from the weeds in the pots were processed and to determine the control rates. The accuracy of analytical methods and image processing method were discussed. The LPG doses required for ED90 control level in the controlled environment at 2-4, 6-8 and 10-12 Y were 110, 121 and 191 kg ha-1 based on dry matter at DAT14, while these values were 73, 86 and 124 kg ha-1 in field conditions. However, visual evaluations suggested much lower (30-40 kg ha-1) gas doses to be applied at 2-4 Y. As a result of image processing, it was found that 30, 90 and 94 kg ha-1 at the 2-4, 6-8 and 10-12 Y periods were sufficient for 90% control rate, respectively. Since the effect of weed trunk could not be taken into account sufficiently in image processing, the required LPG doses were much lower than those of dry matter analysis. In addition, due to leaf overlaps and curls, the rate of error in the leaf area calculation was high in image processing. It was concluded that image processing method used in this study was inappropriate and was not as accurate as analytical methods.