The invention discloses a multi-spectral remote sensing image change detection method based on a spectral reflectivity neighborhood difference chart and neighborhood probability fusion. The method mainly comprises the steps that 1, a registered two-time-phase multi-spectral image set at the same area is input and is subjected to Wiener filtering noise reduction and normalization processing; 2, the processed image set is converted into a relative ground object spectral reflectivity image set; 3, the spectral reflectivity change module value is calculated, and a spectral reflectivity change module value chart is obtained; 4, the spectral reflectivity angle value is calculated, and a spectral reflectivity angle value chart is obtained; 5, a neighborhood difference value of the spectral reflectivity is calculated and the neighborhood difference chart is obtained; 6, clustering is carried out on the spectral reflectivity change module value chart, the spectral reflectivity angle chart and the neighborhood difference chart, and a binary image is obtained; 7, the obtained binary image is fused based on the neighborhood probability, and a change detection result chart is obtained. Manual participation is not needed, the detection precision is high, and the method can be used for the fields such as urban expansion monitoring, forest and vegetation change monitoring and the like.