Cooperative relaying is an enhancing technique for Device-to-Device (D2D) communications, which can significantly expend the communication distance and increase the capacity of D2D links. However, it is challenging to balance the probing time for selecting the optimal relay node and the effective transmission time, both of which will affect the overall throughput of D2D transmissions. In this work, we first build a tractable model for a multi-cell heterogeneous D2D network based on stochastic geometry to evaluate the coverage probability of three location-aware relay selection schemes, namely, All Relays Selection, Sectorized Relays Selection and Distance Based Selection. Based on this model, we develop a practical channel probing process to implement these relay selection schemes and investigate the corresponding transmission capacity under the probing-transmission tradeoff. Numerical results reveals that Distance Based Selection offers the best performance in terms of both coverage probability and transmission capacity when considering the channel-probing cost.