This paper investigates the optimal number of collaborating secondary users which can minimize the total error rate (i.e., the summation of probabilities of missed detection and false alarm) for cooperative spectrum sensing when the secondary users follow Poisson point process (PPP). The received signal-to-noise ratios (SNRs) at secondary users can vary since the practical locations and the amount of secondary users are various. In order to cope with the diversity of received SNRs, a practical cooperative spectrum sensing system model is proposed based on the generalized likelihood ratio test (GLRT) detector. Firstly, the analytical expressions of the probabilities of false alarm and detection are derived. Secondly, an efficient cooperative spectrum sensing technique is proposed, which only requires the least amount of cooperative secondary users to achieve the minimum total error rate. Furthermore, the analytical results reveal the chosen optimal numbers of collaborating secondary users can achieve the best spectrum sensing performance.