In order to improve the assembly performance of Optical Burst Switching (OBS) networks, an adaptive assembly strategy for the edge node of OBS was presented in this paper. To be different from the traditional assembly strategy with a fixed threshold system, the proposed strategy measured the network traffic flow and chose a suitable assembly threshold dynamically by introducing a cross counter.Moreover, the step length of the assembly thresshold was modified by comparing the type of assembly threshold with the former one. Simulation results show that the strategy can reduce the burst loss probability by 69.07% and 38.89% as compared with the existing assembly strategy. The edge node hardware structure for a OBS testing bed was designed and realized,and the functions of assembly and schedule based on the hardware structure were tested, meanwhile, the prospective goal was achieved. Results show that the novel strategy not only can adjust the assembly threshold according to the traffic flow, but also can have lower burst loss probability under a strong self-similar case.