Stereo Image Acquisition System Using Optical Axis Movable Liquid Crystal Lens
Wan Chaojie,Liu Zhiqiang,Xu Luhan,Li Huihai,Ye Mao
DOI: https://doi.org/10.3788/aos221402
2023-01-01
Acta Optica Sinica
Abstract:Objective Stereo images are usually acquired by changing the position of a single camera in the scene or by using two or even more cameras fixed at the same platform. Multi-camera systems are large and costly, and due to slight differences between each camera in terms of focal length, zoom level, camera gain, and so on, there are inevitable intensity differences between the matching points of the stereo images. Researchers have adopted methods combining a single camera and several optics devices to achieve stereo vision, so as to avoid this problem. The core function of the optics devices is to develop a single camera system with different imaging views. Various optical components have been reported in studies, for example, by rotating a flat glass plate or some plane mirrors placed in front of a single camera, or using a single camera pointing at a biprism or some multiple parabolic mirrors. However, optics rotation systems need to address the accuracy of mechanical movement, and biprism systems need to solve the problem of how to get the same image size. Parabolic mirror systems using multiple curvatures involve complex mirroring mechanisms, and the mirrors with multiple curvatures make the system difficult to be compact. Therefore, it is of research value and significance to seek more direct methods to realize a single-camera stereo vision system with a simpler and more compact structure. In this study, a single-camera stereo image acquisition system using an optical axis movable liquid crystal lens is presented. The optical axis position of the liquid crystal lens can be controlled by adjusting the voltage, and thus the stereo vision in a single camera can be achieved. Although it is consistent with the purpose of multi-view imaging through rotation, mirror reflection, biprism, etc., the mechanism of the designed system is simple. It allows the system to be used without mechanical movement during image acquisition, so as to reduce the complexity of the system. The liquid crystal lens is thin and light, and fits closely to the camera lens, which makes the system compact and enables a low cost. Methods The system consists of a fixed camera module and a liquid crystal lens with a polarizer attached. First, the structure of the liquid crystal lens is described, and a polarized interference optical path is built to analyze optical axis movement properties, including the magnitude of the motion at the corresponding drive voltage as well as the aberration and optical power. Then, the effect of the optical axis position change of the liquid crystal lens on the overall system is analyzed, and the relationship between disparity and depth is derived through the pinhole camera model. Finally, the system is used to acquire stereo images, and the disparity is calculated by the optical flow algorithm. The depth information of the scene is inferred from the disparity information in the close-range experiment, and error analysis is performed. In addition, the problems of other disparity acquisition algorithms in the designed system are illustrated. Results and Discussions The proposed single-camera system using an optical axis movable liquid crystal lens is proved to have a stereo vision effect. The images in this paper are acquired by using the horizontal movement of the optical axis of the liquid crystal lens while the single camera is fixed. The images have consistent direction of pixel movement and variation in the amount of pixel shift depending on the object distance. At this point, the disparity is the same magnitude as the optical flow. Therefore, the difference in the amount of pixel movement by using the optical flow algorithm is calculated, and dense disparity information is obtained [Fig. 9(b)]. It indicates that the system can capture disparity information about the scene (Fig. 10). Depth of objects is inferred back from the disparity in the close-range experiment (Table 3). The baseline distance in this system is small, and conventional stereo vision algorithms such as semi-global matching (SGM) and sum of absolute differences (SAD) are unable to acquire dramatic disparity information on the images acquired by the system (Fig. 12). According to the binocular distance measuring principle, as the baseline distance becomes greater, higher accuracy and longer distance that can be measured. It has been proven that the designed system works well, and it can acquire dense disparity information through the optical flow algorithm and allow the depth measurement of objects at close range with some accuracy. Conclusions In this study, a stereo image acquisition system is proposed. The system consists of a fixed camera module and a liquid crystal lens with a polarizer attached. We adjust the voltage to move the optical axis of the liquid crystal lens, capture images, and use optical flow algorithm to obtain disparity information. In addition, we analyze the effect of the optical axis movement of the liquid crystal lens on the overall optical axis of the system, and derive the relationship between the disparity and the depth. In the experiment, the feasibility of the system is demonstrated by verifying the existence of the disparity in the acquired images, and depth acquisition of close-range objects is performed. Experimental results show that the optical axis movement function of the liquid crystal lens can be used to move the overall optical axis of the system to achieve stereo vision. The designed system does not require any mechanical movement, and features a simple and compact structure and a low cost. Therefore, it provides a new method for the acquisition of stereo images.