Multiclass Classification on Soybean and Weed Species Using a Novel Customized Greenhouse Robotic and Hyperspectral Combination System

Abstract

Soybean production is greatly affected by different types of weeds such as horseweed, kochia, ragweed, redroot pigweed, and waterhemp in the Midwest region of U.S. Identification of the soybean plants and the weeds are crucial to control the weed in precision agriculture. The objective of this study was to classify soybean plants and 5 weed species where a hyperspectral imaging camera with a spectral range from 400-1000 nm was used to acquire the images. To acquire the HSI images, a customize robotic hyperspectral data collection scanning platform was developed and used in the greenhouse. A total of 983 hyperspectral data cubes were captured from the greenhouse environment (n=252, soybean; n=731, weeds). Spectral information was extracted from the collected images and then a classification model was developed by applying partial least squares regression (PLSR) analysis. To construct the calibration and validation data set, the images were divided into 70% and 30% ratio for model training and testing, respectively. Eight types of data preprocessing techniques including mean normalization, maximum normalization, range normalization, multiplicative scatter correction (MSC), standard normal variate (SNV), Savitzky–Golay first derivatives, Savitzky–Golay second derivatives, and data smoothing were explored individually and accepted the best preprocessing method based on the highest performance on calibration and validation results. The results showed maximum validation model performance was found 86.2% by applying Savitzky–Golay second derivatives preprocessing method. The most important wavelength information was evaluated from beta coefficient developed using the same preprocessing method. Finally, chemical images were generated using best performer model to identify the soybean plants from weeds. The generated images showed a significant difference in chemical composition between soybean and weed plants at 443 nm, 553 nm, 633 nm, 743 nm, and 968 nm. The correlation between these peaks and chemical components of the plants are α-caroteniod, anthocyanin, chlorophyll, and moisture respectively. This study shows promising results for the application of HSI in weed control system for soybean and relevant weeds identification in precision agriculture applications.

Keywords: greenhouse robotic system, Hyperspectral imaging, Precision Agriculture, weed identification