Dynamic multi-objective optimization problems are those whose Pareto Optimal Solution (PS) or Pareto Optimal Frontier (PF) changes with the environment or time. Existing methods often use experience to guide and predict future solutions. Most of them ignore the independent and identically distributed characteristics of solutions at different times, and they cannot simultaneously solve the problems of regular and irregular changes. This paper proposes a dynamic multi-objective optimization algorithm (ITDM_DMOEA) based on individual transfer and diversity maintenance. The proposed method uses transfer learning technology to select representative solutions to migrate to the new environment by reusing experience and reference vectors to meet the independent and identically distributed conditions. Secondly, the solutions generated by transfer are merged with the randomly generated solutions, and qualified individuals are selected as seed populations by fitness values. The approximate PS and PF at the new time are obtained by the static optimization algorithm non-dominated sorting genetic algorithm. ITDM_DMOEA is tested on 22 benchmark problems and compared with state-of-the-art DMOEAs. Experimental results show that the algorithm can effectively solve dynamic multi-objective problems.