Objective: In order to solve the problem that the spectral image recognition technology is susceptible to interference from similar types, resulting in deviations in the identification of the type and quality of the measured object. Methods: An independent dual-channel spectral acquisition system for visible light and near-infrared light was designed. By controlling the spectral range and spectral resolution of different characteristic regions, the rapid acquisition of absorbance at the characteristic wavelength position was achieved. The function expression form of spectral change and sample quality was established. According to the spectral distribution characteristics of the test sample, the appropriate characteristic wavelength position was selected, and the calculation basis of the species and quality parameters was given by principal component analysis. Results: The visible and infrared spectra of four common nectarine samples were obtained by CM-25D spectrometer and FT-NIR spectrometer. Compared with the traditional linear proportional analysis method, the average recognition rate of the algorithm was 96.7%, the normalized quality coefficient was 0.892, and the recognition ability was enhanced. Conclusion: The dual-channel spectrum acquisition hardware structure and the partial least squares algorithm based on weight assignment can better classify and identify nectarine varieties with similar spectral characteristics.