For the uplink resource allocation problem of cognitive heterogeneous cellular networks， a resource allocation algorithm based on bandwidth and power constraints is proposed and solved using an improved swarm intelligence algorithm. Based on the characteristics of cognitive radio technology， a range of bandwidth and power allocation values for cognitive home users are derived， and more resources are allocated to other users under the guarantee of satisfying user quality of services （Quality of Services，QoS） to enhance the transmission demand of users in the network and relieve the uplink access load of the network. To address the shortcomings of the bottleneck swarm algorithm such as low convergence accuracy and slow convergence， the crazy operator and dynamic elite learning factor are introduced into the leader and follower， respectively， to improve the algorithm's optimality-seeking efficiency and optimality-seeking accuracy. The improved Salp swarm algorithm is solved for the resource allocation algorithm based on bandwidth and power constraints. Simulation experiments show that the resource allocation algorithm with the introduction of bandwidth and power constraints can be effective in improving network performance， and it can effectively improve system efficiency and user access fairness under the condition of ensuring user QoS.