Detection of Pesticide Residues in Banana Peel Based on Hyperspectral Imaging and Deep Forest Algorithms
Year: 2025
Type: conference-paper
Abstract: This study proposes a rapid detection method for pesticide residues in banana peel using hyperspectral imaging technology combined with a deep forest (DF) algorithm. Spectral data of banana peel samples—treated with dichlorvos, propargite, lambda-cyhalothrin, and imidacloprid, as well as untreated controls—were first collected using a hyperspectral imaging system. Subsequently, three dimensionality reduction techniques, namely Random Frog Leaping (RFL), Competitive Adaptive Reweighted Sampling (CARS), and Successive Projection Algorithm (SPA), were employed to preprocess the hyperspectral data. On the reduced datasets, support vector machine (SVM), random forest (RF), and DF models were established and comparatively analyzed. Experimental results demonstrated that all three dimensionality reduction algorithms effectively reduced data dimensionality and improved modeling efficiency, with RFL showing the most significant enhancement in model performance. Among the classifiers, the DF model outperformed SVM and RF in both accuracy and stability for pesticide residue identification. Notably, when combined with RFL, the DF model achieved an overall accuracy (OA) of 93.28% and a Kappa coefficient of 0.92. These findings confirm that integrating hyperspectral imaging with DF algorithms enables rapid and accurate detection of pesticide residues in banana peel, offering an efficient and convenient approach for agricultural product quality inspection and food safety assurance.
Authors: Rongchang Jiang, Dandan Qu, Ruocheng Yin, Shuren Lu, Shijie Xie
Institution: Jiangsu College of Engineering and Technology
Language: English
Subfield: Analytical Chemistry
Field: Chemistry
Domain: Physical Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W4417504635