In order to maximize the discrete-input continuous-output memoryless channel (DCMC) transmission capacity under certain restriction of modulation scheme in wireless optical multiple input multiple output (MIMO) communication (SM-OMIMO) system with spatial modulation (SM), a novel adaptive power allocation algorithm (PAA) is proposed by the Monte Carlo simulation process. Power allocation coefficients are optimized adaptively for any signal noise ratio (SNR). A model of indoor communication environment is built, and DCMC capacity expression of SM-OMIMO is derived. The distribution principle and procedures of the algorithm are analyzed. Performance of the system is simulated. The effects of several allocation algorithms on DCMC capacity are compared under different combinations of transmits and receivers. Simulation results indicate that, under the condition of low SNR, the system DCMC capacity of system obtained by the proposed adaptive PAA is larger than that obtained by PAA with fixed factor and average PAA, and the proposed PAA is easier to achieve saturation value of DCMC capacity under the condition of high SNR. The results clearly demonstrate the diversity characteristics of the channel. Therefore, the adaptive PAA is capable of improving the transmission rate of SM-OMIMO system.