In order to solve the problems of inconvenient and complexity of the measurement of the crane arm’s welding distortion, the fusion method of support vector machines (SVM) and fuzzy logical theory is adopted to establish the measuring system. In the meanwhile the theory of the genetic algorithm is adopted to optimize the parameters of kernel function and the penalty factor, in the process the error function is defined as the evaluation index, thus parameters of kernel function and the penalty factor are confirmed,σ=1.8 and c=26. The established measuring model is applied to the prediction of welding distortion, meanwhile intelligent image detection method is used to take a measurement of the same prediction point. Contrast results show that the measurement errors between fusion methods proposed in paper and traditional measurement method are under 5%, and accuracy of the measurement method is verified.