varying levels of success. In particular, spectral analysis has been used as a method for discriminating plants from soil (green-from-brown discrimination) (Felton and McCloy, 1992; Brownhill, 2006). Fundamental work presented by Wang et al. (2001) quantifies plant spectral characteristics by using five feature wavelengths and four normalised colour indices for crop/weed (green-from-green) discrimination. The work led to the development of an optical weed sensor capable of detecting wheat from specific weeds under controlled laboratory conditions. More recently, Deng et al. (2014) have investigated the application of Support Vector Machine (SVM), Artificial Neural Network (ANN) and Decision Tree (DT) based classifiers to compare the classification rates for plant spectral measurements taken in the 350–760 nm visible wavelength range and the 350–2500 nm visible and NIR wavelength range. Experimentally, spectral irradiance measurements for the test crop (corn) and weeds (Dchinochloa crasgalli and Echinochloa crusgalli) were performed in the field, using a handheld spectroradiometer, with results showing that the visible light range proved adequate to meet discrimination requirements for the given test plants. An innovative approach presented by Sahba et al. (2006) and Paap et al. (2008) demonstrated
varying levels of success. In particular, spectral analysis has been used as a method for discriminating plants from soil (green-from-brown discrimination) (Felton and McCloy, 1992; Brownhill, 2006). Fundamental work presented by Wang et al. (2001) quantifies plant spectral characteristics by using five feature wavelengths and four normalised colour indices for crop/weed (green-from-green) discrimination. The work led to the development of an optical weed sensor capable of detecting wheat from specific weeds under controlled laboratory conditions. More recently, Deng et al. (2014) have investigated the application of Support Vector Machine (SVM), Artificial Neural Network (ANN) and Decision Tree (DT) based classifiers to compare the classification rates for plant spectral measurements taken in the 350–760 nm visible wavelength range and the 350–2500 nm visible and NIR wavelength range. Experimentally, spectral irradiance measurements for the test crop (corn) and weeds (Dchinochloa crasgalli and Echinochloa crusgalli) were performed in the field, using a handheld spectroradiometer, with results showing that the visible light range proved adequate to meet discrimination requirements for the given test plants. An innovative approach presented by Sahba et al. (2006) and Paap et al. (2008) demonstrated
การแปล กรุณารอสักครู่..