Fringe projection profilometry has been widely used due to its high accuracy, high robustness, and non-contact characteristics. In this paper, we aim to improve the speed and accuracy of the fringe projection profilometry method, especially its performance in jittery environments. In-depth research is conducted. Typical optical 3D sensing technologies mainly include photometric stereo vision, binocular multi-eye stereo vision, time of flight method, laser line scanning method, defocus shape recovery method, and structured light projection method. Structured light projection also includes stripe projection and speckle projection methods. The decoding schemes in fringe projection profilometry are divided into the spatial phase unwrapping method and the temporal phase unwrapping method. The former only requires one phase map to recover the absolute phase, but it relies on the phase values of adjacent pixels, which cannot achieve reliable decoding for discontinuous or isolated objects. The latter projects a series of patterns for decoding, and the absolute phase value corresponding to each pixel is independently calculated, independent of the surrounding pixels. Therefore, theoretically, any shape of the object surface can be unfolded. In the binary fringe method, in order to enhance the contrast of the stripe pattern, binary fringes are used instead of projecting sinusoidal fringes. However, traditional methods require more stripes. For example, the classic method of stripe edge detection requires adding reverse stripes to achieve accurate stripe decoding and positioning. The goal of this paper is not only to reduce the number of stripes but also to effectively improve the accuracy of localization.

Edge localization of binary fringes is a key issue. Song et al. proposed a three-dimensional measurement method based on stripe edge detection. The use of binary stripes instead of sinusoidal stripes as projection patterns greatly enhances the contrast of stripe patterns. At the same time, edge points of binary fringes are detected to reduce interference caused by infrared imaging. Ye et al. combined the stripe edge detection method with near-infrared light to perform a three-dimensional reconstruction of dynamic scenes. However, the stripe edge detection method itself has no resistance to potential jump errors that may occur during the decoding process. To eliminate jump errors, Feng et al. proposed a global codeword correction method that restores continuous and complete point cloud information. By combining stripe edge detection with the global codeword correction method, it is possible to achieve 3D measurements with higher accuracy than traditional phase shift methods. When the measured object experiences shaking, there will be a deviation between the forward and reverse binary fringes, causing the edge points of the two to no longer be in the same position. The stripe edge detection method will calculate the edge points, resulting in a deviation. By taking the periodic ambiguity of gray code order as an example, which is four pixels, the stripe edge detection encoding scheme requires the same number of reverse stripes to be combined. In other words, an additional double of the corresponding reverse stripe projection must be added to solve for edge points. Further, jitter often causes positioning deviation, which leads to errors. That is an important source of error. The performance in a jitter environment will be improved by using a new method in this article. The proposed method does not require specialized projection of reverse stripes corresponding to forward binary stripes and can achieve accurate edge point localization using adjacent images.

The method proposed in this article can effectively eliminate the problem of inaccurate edge point positioning caused by jitter. Furthermore, the effectiveness and accuracy of the method are validated through measurement experiments in jitter scenarios. In the new scheme, only one cyclic reverse stripe pattern needs to be added at the beginning and end of the binary stripe sequence. In the method proposed in this article, the reverse stripe corresponding to the forward binary stripe is not necessary. In addition, achieving accurate edge point positioning is based on adjacent images to obtain the final accurate information. The traditional stripe edge detection method can cause offset in the forward and reverse binary fringes in jitter scenarios, resulting in errors. The new method accurately corrects errors and achieves good results. The experimental results demonstrate that the new method achieves precise positioning of binary stripe edge points through adjacent three frame stripe images. The measurement object is used as the standard to measure the quality of the measurement results. The results are shown in Table 1. The proposed new scheme reduces the number of stripes and uses three adjacent cyclic reverse stripes to locate edges, resulting in more accurate edge points.

In terms of edge localization of binary fringes, this method not only reduces the number of fringes but also effectively resists positioning offset caused by jitter. Compared with the traditional stripe edge detection encoding scheme, in the traditional scheme, a reverse stripe with the same number of stripes as the original encoding is required to obtain the encoding result. The method proposed in this article can achieve more accurate results with fewer stripes, which is significantly superior to traditional methods.