Broad-spectrum optical systems have superior performance. They can obtain more comprehensive and accurate target information and are conducive to enhancing the detection and identification capabilities of optoelectronic equipment. In addition, they have an irreplaceable role in complex environments. However, their design is often difficult, and the current main optical design method is to optimize the selected initial structure, but the initial structure of the broad-spectrum optical system is inefficiently constructed. Therefore, the design cycle is long, and it relies too much on the experience of the designers. In this paper, we explore the design method of broad-spectrum optical systems, analyze the system's confocal and co-image plane conditions from the theoretical level, and focus on the initial structure construction method of broad-spectrum systems based on a genetic algorithm.

Firstly, a multi-band equal focus method based on optical focus matching is proposed to derive the equal optical focus condition by taking a system with two bands and two optical groups as an example, and the system's focal lengths in each band are set to be equal by reasonably allocating the optical focus of each optical group. Then, in order to meet the different wavelengths of the common image plane imaging, the geometric optics formula recursively obtains the combined system of equal image plane conditions. The idea of a genetic algorithm is used to independently construct the optical group composition and structure form of the broad-spectrum optical system and iteratively solve the optimal initial structure, and the selection of the optimal group is based on the experience of the designers, which is mainly considered in the differences in the focal lengths of different wavelengths, the differences in the position of the imaging surface, the distribution of the optical focal lengths, the ratio of the lens diameter to thickness, and the thickness of the edge of the lens, and other aspects. Finally, the selected optimal initial structure is optimized to obtain a broad-spectrum optical system with a small number of lenses, small volume, light weight, and good imaging quality.

For the same target, different focal lengths will lead to differences in the position and size of the image, and with the broadening of the spectral range, the imaging differences increase, which will seriously reduce the imaging quality of the system. The theoretical derivation of this paper obtains the conditions of equal focal lengths and co-image surfaces, which can solve this problem. For the optimal initial structure construction of the transmissive system, this paper independently constructs the optical group composition and structural form of the system by genetic algorithm, while the existing papers use optimization algorithms for the design of reflective systems, such as the design of reflective free-form surfaces. In the initial structure construction process of such systems, the parameters of the incident light and the requirements of the outgoing light are determined, and the algorithm process actually fits the reflective surface according to the laws of geometrical optics under the premise of known incident and outgoing light. In contrast, the transmissive system contains multiple optical groups; the light propagation path inside the system is completely uncertain, and the number of optical groups and the structure form are all unknown. Therefore, the main framework strategy and process of the broad-spectrum initial structure construction algorithm in this paper are completely different. In addition, we generate many possible optimal solution results through the powerful computational ability of the algorithm, which ensures a high probability of occurrence of the optimal solution by the number and effectively prevents the algorithm from falling into the local optimum. The results of the algorithm (Fig. 3) indicate that it can efficiently generate a large number of excellent initial structures of the broad-spectrum system, providing training samples for the later AI-based optical system design.

In this paper, we derive the broad-spectrum co-focal distance and co-image plane equations, determine the focal distance difference and image plane difference fitness function, and establish the genetic algorithm structure parameter variation and material crossover method. In order to verify the feasibility and efficiency of the method, a visible and near-infrared broad-spectrum optical system is designed, and the system has an imaging band of 0.4–1.2 μm and a focal length of 40 mm. The difference in the focal length within the range of the band is less than 0.03 mm, and the imaging quality is good in the broad-spectrum range. The design results show that the genetic algorithm-based broad-spectrum optical system construction method can generate 1024 excellent populations at a time, of which 407 iterative individuals can meet the requirements of the objective function. The appropriate optimal solution is input into the Zemax software for optimization, and a broad-spectrum optical system that meets the requirements can be obtained very quickly. In summary, the proposed genetic algorithm can shorten the design cycle and improve the design efficiency of broad-spectrum optical systems.