With the rapid development of integrated photonics, automatic alignment of fiber-to-waveguide edge coupling has become a key challenge in testing photonic quantum chips. Traditional local optimization methods often get trapped in local optima and converge slowly when they are used for alignment problems that are high-dimensional, strongly coupled, and contain many local optima. To address the high-dimensional global optimization problem in a dual-end coupling system, this paper presents an improved genetic algorithm. The algorithm uses a chromosome structure designed for the coordinates of both ends and introduces a crossover operation to improve search efficiency. This improves both global search and local refinement. This paper realizes active automatic alignment for a 12-dimensional coupling system. The results show that the system can effectively converge to the global optimal coupling position with high consistency. The average alignment time measured in the experiment is 229.2 s and the average coupling loss is about 0.2 dB, which verifies the effectiveness and practicality of the algorithm for high-dimensional coupling alignment problems and provides a feasible solution for high-dimensional global optimization under a single sensing parameter.