In order to quickly find an optimal frequency point in the HF frequency band, a HF bidirectional detection frequency selection algorithm based on the variable Neighborhood particle swarm search (VNSPSO) is proposed in combination with the broadband spectrum sensing technology. The existing detection frequency selection algorithms are difficult to meet the needs of realtime frequency selection because they are mainly based on the average signaltonoise ratio of the frequency points to evaluate and select the best, and the smallscale random fading characteristics of the shortwave channel is left out of consideration. According to the correlation characteristics of largescale fading, the initial detection frequency set is obtained by the maximum separation method to divide the correlation neighborhood. According to the characteristics of mass selective fading of neighborhood internal frequency points, particle swarm optimization algorithm is used to search neighborhood internal frequency points and to obtain the local optimal solution. The global optimal solution is obtained by transforming the neighborhood. The simulation experiment shows that when “The fastest speed” is used to build the chain, compared with VNSRS, AASS and RSS, MTOBC of the VNSPSO algorithm reduces by 17.1%, 18% and 85.5% respectively. When the CPOS=0.9, the chain building time decreases by 2.5%, 42.6% and 81.7% respectively, the time for establishing a passable link is shortened. When “optimal frequency point” is used to build the chain, the MTOBC of VNSPSO algorithm reduces by 11%, 12.5%, and 45% respectively, compared with VNSRS, AASS, and RSS algorithms. When the CPOS=0.9, the chain building time reduces by 22.2%, 22.4%, and 444% respectively, and the optimal frequency point can be found in a short time.