Objective To achieve real-time, non-destructive, and high-precision monitoring of multiple quality indicators of kiwifruit during storage and to address the problems of strong destructiveness, obvious lag, and the difficulty of a single data source and model to balance comprehensiveness and accuracy in conventional detection methods, a kiwifruit quality testing method integrating wireless sensor network (WSN), hyperspectral imaging, and Harris hawk optimization (HHO)-optimized deep extreme learning machine (DELM) is proposed.Methods Firstly, a WSN and hyperspectral imaging collaborative acquisition system is established to obtain kiwifruit storage environment parameters (temperature and humidity), fruit surface sensing data (firmness and skin color), and hyperspectral image data. Secondly, the spectral data are preprocessed by the standard normal variance (SNV) combined with the first derivative, and the competitive adaptive re-weighted sampling (CARS) algorithm is employed to extract characteristic wavelengths for feature dimensionality reduction. Finally, a multi-source data fusion model for kiwifruit quality testing is established by optimizing the DELM with the HHO algorithm.Results The HHO-DELM model on the test dataset achieves correlation coefficients of 0.972, 0.915, 0.903, and 0.926 for soluble solids content, firmness, acidity, and color difference (e value) in the prediction dataset, with root mean square errors (RMSEs) of 0.486 °Brix, 12.980 N/cm², 0.062, and 0.023, respectively, and residual prediction deviation (RPD)>3.0 for all indicators. The established method reaches the classification overall accuracy of 99.3% (Kappa=0.991), significantly outperforming DELM, partial least square regression (PLSR), support vector machine (SVM), and random forest (RF). Under different storage conditions (cold and room temperature), the method shows the prediction accuracy fluctuations ≤2.5%, demonstrating excellent robustness.Conclusion This method integrates the real-time monitoring of WSN, the multidimensional information acquisition of hyperspectral imaging, and the high-precision prediction of HHO-DELM, enabling non-destructive testing of multiple quality indicators of kiwifruit throughout the storage period, thus meeting industrial requirements for rapid on-site detection and dynamic monitoring.