Core loosening is one of the main mechanical faults of high voltage shunt reactors under long-term vibration effect. The defect diagnosis by single-point vibration signal analysis method has some drawbacks such as difficult selection of test points and non-typical data. The above problems can be effectively avoided by using vibration sensor array. This paper introduces the vibration mechanism of the core of shunt reactor based on mass-spring model. With the vibration test platform of high voltage shunt reactor，a test scheme of vibration morphology of high voltage shunt reactor tank is established， and the vibration morphology of high voltage shunt reactor tank surface under each frequency component is obtained. The influence of the voltage and core loosening on the vibration morphology of high voltage shunt reactor tank surface is analyzed. A method based on texture feature extraction in image morphology is presented. The feature extraction algorithm of the vibration intensity symbiosis matrix is applied to a 10 kV shunt reactor model with a fixed core and different loosening states to verify the effectiveness of the proposed algorithm.