In order to improve the detection performance of sensing users for primary users with weak signals in the cognitive radio environment, a distributed optimal collaborative detection algorithm with multiple sensing users is proposed. In this algorithm, with the known probability dist.ributions of primary user＇ s signals and noises, both the optimal fusion rule of the detection system and the optimal decision threshold of each sensing user are obtained via the sequential quadratic programming in the optimization theory, which makes the collaborative detection probability maximum. Being independent of the detection form of the receiver, the proposed algorithm possesses high convergence rate of numerical solution. Simulated results indicate that, with five sensing users, 255 cumulative correlated symbols, 10% false alarm of the fusion center and the same signal-to-noise ratio of all sensing users, the proposed algorithm is superior to the existing best detection algorithm based on the fixed fusion rule in both the additive white Gaussian noise and the fiat Rayleigh fading channels ,with an improved collaborative detection probability by 13% and 5% , respectively.