The paper introduces that railway track online profile inspection based on machine vision is influenced by vehicle vibration. It presents a project that eliminates the influence of vehicle vibration by stabilized device. This article designs based on singleaxis gyro stabilized device, and modeling analysis was carried out on stability control circuit. What’s more, single neuron PID adaptive control based on RBF neural network online identification has come up. Simulation manifests that compared with traditional PID this algorithm can solve the problem of inaccurate object model effectively, and it has the characteristic of rapid response speed and high stability. This provides theory basis for highprecision stabilized device.