Chaos possesses outstanding pseudo - random properties. The chaotic time series fits well for the initialization of food sources in the artificial bee colony algorithm. However, the distribution of most chaotic mappings is not homogenized, which lowers the search efficiency and accuracy of the traversal algorithm. Hence, this paper puts forward a novel homogenization-based optimization method for population initialization of improved chaotic artificial bee colonies. According to the principle of maximum entropy, Logistic chaotic mapping is optimized to achieve homogenization. Subsequently, the randomness of the generated homogenized time series is checked through entropy spectrum analysis and NIST randomness test. Moreover, a dynamic grouping strategy based on fitness is designed to balance the exploration and development phases of the algorithm. Furthermore, experimental simulations are conducted on 18 standard test functions. It is also compared with other optimization algorithms in terms of convergence curve and optimization result. And related algorithms are appropriately introduced into the material flow distribution to seek the shortest path. The results indicate that the proposed algorithm enables the initial value to be more homogeneously distributed in the search space, thus improving the global pioneering of the algorithm. Through dynamic grouping of the population, the fine - grained search ability is enhanced.