For soil moisture stress detection of maize, the physiological characteristics indicators are commonly used, but such methods can affect the growth of maize plants. To solve this problem, a maize soil moisture stress predictive model based on multiview stereo vision and support vector machine (SVM) with error correcting output code (ECOC) was proposed. Firstly, an RGB camera was used to obtain three maize images which was at -30°, 0° (maize leaf expansion plane) and 30° during the jointing stage. The obtained images were segmented in the HSV color space to extract the whole maize plant. The discrete areas were extracted and removed simultaneously by calculating the size of the connected domain and retaining the largest connected domain. Morphological dilating was used to smooth the edges of the extracted maize leaves and fill the holes of leaf, and the edge information was detected by using the Scharr filter. Then, two maize cloud models of -30°~0° and 0°~30° were established based on the stereo vision of speeded up robust features (SURF). In the process, the fast library for approximate nearest neighbors (FLANN) and random sample consensus (RANSAC) were used to reduce the error matching, and the final feature point matching accuracy was 98.95%. The iterative closest point (ICP) was used to merge the two maize cloud models data into the same coordinate system, and the registration error was less than 0.01mm. The cloud skeleton was extracted by L1median method. Finally, the parameters, including internode height, leaf length and plant height were extracted from the maize plant skeleton, and the water stress prediction model for single parameters and soil moisture stress ECOC-SVM predictive model were established. The results showed that the leaf length, the internode height and the daily growth of maize plant were significantly linearly correlated with the degree of moisture stress. In this research, the above three parameters were taken respectively as independent variables and the soil moisture content as dependent variable to establish the moisture stress predictive models. The correlation coefficients were 0.8922, 0.8928 and 0.8176, and the RMSE were 2.92%, 2.53% and 2.76%. In order to improve the prediction accuracy, a maize soil moisture stress predictive model of ECOC-SVM was established using above three maize parameters as the characterized vector. The prediction accuracy of the test set was 93.33%, showing that the accuracy of this model was very high. When the maize was at jointing stage, the predicted value of soil moisture content can be obtained from a single parameter maize water stress prediction model, and the degree of moisture stress on maize can be predicted by the multiparameter ECOC-SVM model. The research result can provide technical support for accurate access to agricultural information.