The invention relates to a support vector machine-based method for correcting measurement voltage values of insulator detection robots. The method comprises the following steps of: simulating insulator conditions with different low values and zero values according to an insulator distribution parameter model and an insulator detection robot equivalent circuit model, respectively calculating an insulator string voltage distribution value and when a robot does not exists and an insulator string voltage distribution value measured by the robot at each insulator, and taking the two groups of dataas samples to form a sample library; selecting a sample set, carrying out training by utilizing a support vector machine model, carrying out correction to obtain practical voltage distribution, and completing model training through model parameter optimization; and correcting measurement data by utilizing the trained model, so as to obtain real insulator string voltage distribution. According to the method, a complicated nonlinear relationship between measured voltage values and practical insulator string distribution voltage can be obtained through sample data training and learning under typical defect conditions, so as to correctly correcting measured voltage values under normal conditions and different low/zero-value insulation conditions.