As the precondition of dynamic spectrum access, spectrum sensing (SS) plays an important role in the spectrum utilization in cognitive wireless mesh networks (WMNs). However, the accuracy and efficiency of SS is always significantly compromised by the accompanying interchannel interference (ICI). Aiming to address this problem, this paper proposes a non-coherent ICI cancellation method in the cooperative SS. The proposed SS consists of three parts: the determination of optimal threshold, the ICI cancellation, and the algorithm of control channel allocation (ACCA). The derived optimal threshold is able to minimize the sum of the probabilities of false alarms and missed detection. The ICI cancellation is implemented indirectly by solving a nonhomogeneous linear equation set with a coefficient matrix depending only on the distances between the primary transmitters and the cognitive secondary users. The ACCA is able to utilize multiple in-band control channels rather than a single common control channel; thus, it can decrease the sensing delay. Simulations show that the proposed SS can efficiently cope with the ICI and achieve much better performance than the conventional methods in the low signal to interference-plus-noise ratio range.