Advances in weed identification using hyperspectral imaging: A comprehensive review of platform sensors and deep learning techniques

Abstract

Hyperspectral remote sensors are emerging as valuable technology for identifying weeds in field crops. These sensors can acquire high-resolution images, both spatially and spectrally, which are crucial for early weed identification. The objective of this paper is to provide a comprehensive review on studies that have utilized hyperspectral remote sensing for weed identification in field crops. The review further explored the weed identification accuracies of these remote sensors. To gather the relevant information, a literature search was conducted across three academic journals (Science Direct, Web of Science, and Google Scholar) using a search criterion. The review summarized key findings as follows; (a) remote sensors such as the Specim ImSpector V10E, ASD FieldSpec, and Resonon Pika II are the widely used remote sensors for both ground and aerial data acquisition, (b) majority of the research utilized ground-based mounted platform for data acquisition compared to unmanned aerial platforms, (c) preprocessing hyperspectral data is critical for weed identification. Techniques like image calibration, standard normal variate, multiplicative scatter correction, Savitsky-Golay smoothing, derivatives, and features selection are among the most used techniques, (d) traditional machine learning models namely support vector machines (SVM), partial least square discriminant analysis (PLS-DA), maximum likelihood classifiers (MLC), and random forest (RF) are the widely employed classifiers for weed identification, (e) the application of deep learning technique, namely convolutional neural networks (CNNs) are limited, but its application demonstrated superior performance accuracies compared to traditional machine learning models. In conclusion, this review seeks to provide valuable insights into the latest developments in hyperspectral remote sensing for weed identification.

Keywords

• Deep learning
• Data preprocessing
• Hyperspectral imaging
• Machine learning
• Remote sensors
• Weed detection