This paper presents an adaptive real-time spectrum data compression and recovery method for radio monitoring, its purpose is to reduce the data storage space and the network bandwidth occupied by transmission without affecting the subsequent analysis and application. Considering the similarity between spectrum data, we use correlation coefficients and bitmap similarity to measure them, and then replace all the original spectrum with a small amount of typical spectrum to achieve the purpose of compressing the original spectrum for storage and compression. The experimental conclusions show that the method can automatically adapt to various radio frequency bands and achieve better compression effects.