At present, nearly all optimization methods used for optical thin film design can seek only one object, which is difficult to meet the complex needs of optical thin films. So, the multi-objective optimization model of optical thin film and a new multi-objective genetic algorithm (DMOGA) are proposed. In DMOGA, a method based on non-dominated relationship is used in choosing operator. Solution with qualified distributivity and approximation by shrink of the population of non-dominated set is achieved based on dynamic crowding distance and the algorithm parameters are adjusted dynamically to improve the efficiency of algorithm. Some examples of optical thin film designs are solved by DMOGA. The results show that DMOGA is a useful algorithm for multi-objective problems and may have a bright future in optical thin film design.