Cyperus esculentus (yellow nutsedge) is one of the world’s worst weeds as it can cause great damage to crops and crop production. To eradicate C. esculentus, early detection is key—a challenging task as it is often confused with other Cyperaceae and displays wide genetic variability. In this study, the objective was to classify C. esculentus clones and morphologically similar weeds. Hyperspectral reflectance between 500 and 800 nm was tested as a measure to discriminate between (I) C. esculentus and morphologically similar Cyperaceae weeds, and between (II) different clonal populations of C. esculentus using three classification models: random forest (RF), regularized logistic regression (RLR) and partial least squares–discriminant analysis (PLS–DA). RLR performed better than RF and PLS–DA, and was able to adequately classify the samples. The possibility of creating an affordable multispectral sensing tool, for precise in-field recognition of C. esculentus plants based on fewer spectral bands, was tested. Results of this study were compared against simulated results from a commercially available multispectral camera with four spectral bands. The model created with customized bands performed almost equally well as the original PLS–DA or RLR model, and much better than the model describing multispectral image data from a commercially available camera. These results open up the opportunity to develop a dedicated robust tool for C. esculentus recognition based on four spectral bands and an appropriate classification model.

Jimson weed (Datura stramonium L.) is a toxic weed that is occasionally found in fields with common bean (Phaseolus vulgaris L.) for the processing industry. Common bean growers are required to manually remove toxic weeds. If toxic weed plants remain, the standing crop will be rejected. Hence, the implementation of an automatic weed detection system aiding the farmers is badly needed. The overall goal of this study was to investigate if D. stramonium can be located in common bean fields using an unmanned aerial vehicle (UAV)-based ten-band multispectral camera. Therefore four objectives were defined: (I) assessing the spectral discriminative capacity between common bean and D. stramonium by the development and application of logistic regression models; (II) examining the influence of ground sampling distance (GSD) on model performance; and improving model generalization by (III) incorporating the use of vegetation indices and cumulative distribution function (CDF) matching and by (IV) combining spectral data from multiple common bean fields with the use of leave-one-group-out cross-validation (LOGO CV). Logistic regression models were created using data from fields at four different locations in Belgium. Based on the results, it was concluded that common bean and D. stramonium are separable based on multispectral information. A model trained and tested on the data of one location obtained a validation true positive rate and true negative rate of 99% and 95%, respectively. In this study, where D. stramonium had a mean plant size of 0.038 m² ($\sigma$ = 0.020), a GSD of 2.1 cm was found to be appropriate. However, the results proved to be location dependent as the model was not able to reliably distinguish D. stramonium in two other datasets. Finally, the use of a LOGO CV obtained the best results. Although small D. stramonium plants were still systematically overlooked and classified as common bean, the model was capable of detecting large D. stramonium plants on three of the four fields. This study emphasizes the variability in reflectance data among different common bean fields and the importance of an independent dataset to test model generalization.