The occurrence of weeds is one of the main factors limiting agricultural productivity. Studies on new techniques for the identification of these species can contribute to the development of proximal sensors, which in the future might be coupled to machines to optimize the performance of species-specific weed management. Thus, the objective of this study was to use near-infrared (NIR) spectroscopy and multivariate analysis to discriminate three morningglory species (Ipomoea spp.). The NIR spectra were collected from the leaves of the three weed species at the vegetative stage (up to five leaves), within the spectral band of 4,000 to 10,000 cm^–1. The discrimination models were selected according to accuracy, sensitivity, specificity, and Youden's index and were analyzed with a validation data set (n = 135). The best results occurred when the selection of spectral bands associated with the use of preprocessing was performed. It was possible to obtain an accuracy of 99.3%, 98.5%, and 98.7% for ivyleaf morningglory (Ipomoea hederifolia L.), Japanese morningglory [Ipomoea nil (L.) Roth], and hairy woodrose [Merremia aegyptia (L.) Urb.], respectively. NIR spectroscopy associated with principal component analysis and linear discriminant analysis (PC-LDA) or partial least-squares regression with discriminant analysis (PLS-DA) can be used to discriminate Ipomoea spp.