Spectrum fragmentation has become a primary factor that affects spectrum utilization in elastic optical networks. To reduce the spectrum fragmentation rate, effective measurement of the spectrum fragmentation is necessary. Thus, in this study, a time-frequency fragmentation-aware model is developed based on a detailed analysis of the generation mechanism and probability of fragmentation resources, and a spectrum allocation algorithm is proposed to select the spectrum with the least degree of fragmentation for connection requests. Simulation results show that compared with the contrast algorithm, the proposed algorithm can reduce the service blocking rate and improve spectrum utilization while effectively reducing the spectrum fragmentation rate.