With the rapid development of wine market, a large number of Chinese high quality wine has been affected by inferior wine. The existence of fake inferior wine not only affects quality wine brand in China, will also do a certain harm to human body. Water adulteration in wine is the most common means of making fakes, therefore, study of wine water adulteration detection method has attracted more attention from the researchers both at home and abroad. Compared to traditional sensory assay methodor physical and chemical testing methods operated in laboratory, visible/near infrared spectral analysis technology is more suitable for rapid detection of wine quality with thequickness, high efficiency, non-destruction and non-contactfeatures. In order to detect the wine water blending problem rapidly and accurately, based on the visible/near infrared spectral analysis technology, this paper constructed a spectral absorption Depth Index (DI) to reflect the water degree blended in wine, and gave the wine mixing water inversion model based on DI Index to estimate the water content. First, this paper chosethree kinds of wine including the Changcheng cabernet wine (CC), Zhangyu cabernet wine (ZY) and Xiaocabernet wine(XA)to create 18 wine samples with 0% pure wine (no water), 4%, 7.7%, 11.1%, 7.7% and 17.2% of distilled water in the three kinds of wine respectively, and to create other 6 wine samples with 0%, 20%, 40%, 60%, 80%, and 90% of distilled water in Changcheng wine. So there were totally 24 wine samples with different ratios of distilled water. Then, the wine spectral data were sampled using the PSR-3500 portable features spectrometer. After the preprocessing of the S-G filtering, special wavelength choosing, and continuum removing of the original spectral data, the visible/near infrared spectral features of wine samples were analyzed, anda spectral absorption depth Index (DI) of wine with distilled water was constructed using the stable spectral absorption property at 837 nm. In order to improve the robustness of DI index, the mean value of the spectral reflectance values near 837 nm small neighborhood was adopted. Finally, the wine mixing water inversion model based on DI index was created using the quadratic polynomial fitting method. To validate the inversion estimate model of the wine with water, the DI index of Changcheng cabernet wine was used, and seven samples were chosen as the prediction set, and the other four samples were chosen as test set in the experiment. Experimental results showed that the precision of R square value of the model is up to 0.999 2 with the quadratic polynomial fitting method, and the average relative error between the estimates of the model and the real value is 0.042 5. Experiments showed that the inversion estimated model based on DI index can not only identify whether the wine blended with water, but also make a quantitative analysis of the water content in wine. DI index was simple, and the DI index can reflect the water degree of different brands of wine. This study may provide a scientific basis for the design and development of low-cost and handheld portable spectrometers for wine detection, further promoting visible/near infrared spectral analysis technology in the quality detection of wine or other relative field.