3D Model development

The creation of photo-realistic 3D models is a critical step in the digital preservation of cultural heritage assets. Using a multi-camera setup and multi-image matching technique, this study presents an image-based 3D modelling pipeline that does not require any markers on or surrounding the object to be used in the modelling process. Digital single lens reflex (DSLR) cameras having invariant relative orientations are used in conjunction with a number of other cameras.

Instead of doing photo-triangulation after picture acquisition, calibration is carried out in order to estimate the external orientation parameters of the multi-camera system, which can then be processed completely automatically using coded targets, saving time. The calibrated orientation parameters of all cameras are applied to photos taken with the same camera combination, regardless of which camera was used. Therefore, while doing multi-image matching for surface point cloud generation, the orientation parameters will remain the same as they were when the calibration results were obtained even when the target has changed. Once the entire system has been calibrated and the software has been flawlessly linked, the entire 3D modelling process may be executed totally automatically on the basis of this invariant characteristic. A number of experiments were carried out in order to demonstrate the practicality of the suggested system. A human being, eight Buddhist statues, and a stone sculpture are among the images that have been discovered. The results for the stone sculpture were compared to a reference model acquired using an ATOS-I 2M active scanner. The results for the stone sculpture were obtained using multiple multi-camera configurations. The best result has an absolute accuracy of 0.26 mm and a relative accuracy of 1:17,333, with the absolute accuracy being the most accurate. It highlights the practicality of the suggested low-cost image-based 3D Modeling Services, as well as its applicability to a large number of antiquities preserved in a museum setting, in this case. Photo-realistic 3D models of close-range objects are useful for a variety of applications including cultural heritage documentation, human face and torso modelling, industrial reverse engineering, and other fields of study.

Historically, close range photogrammetry has been the most widely used method of cultural heritage documentation, dating back to 1885. As laser scanner technology has progressed, there has been increased interest in the application of terrestrial laser scanning (TLS) for heritage documentation. Despite the fact that TLS can achieve great levels of geometric detail with high degrees of precision, considerable labour, skill, and time are still required at the stages of data collecting and data processing . The TLS data, on the other hand, is incapable of displaying textural information on its own. A digital camera image must be integrated with and registered with the computer. Not only does this increase the cost of data processing, but it also increases the complexity of the data processing pipeline. For terrestrial 3D modelling, it turns out that image-based modelling is the most cost-effective method that is also the most versatile, portable, and commonly utilised. Among other things, it is widely used not only for close-range applications, but also for airborne  and spaceborne images. When compared to laser scanning, it has a number of advantages, including faster data capture, extensive texture information, knowledge of the measured locations, and the ability to perform measurements at any time, even after the target has been destroyed. For scientific research purposes, the texture is the most significant of them since it allows you to reconstruct a photo-realistic 3D model or a true 3D model that is suitable for future scientific investigation.

It is possible to find in the literature detailed comparisons of current technologies for terrestrial 3D modelling, such as image-based rendering (IBR) and image-based modelling (IBM), and range-based modelling. According to , there is no one modelling technique that can meet the requirements for varied close range applications with respect to geometric correctness, amount of detail, degree of automation (portability), flexibility (flexibility), photo-realism (photorealism), cost, and efficiency. The integration of several approaches and data sources is therefore advocated, particularly for the modelling of large-scale cultural sites , which is particularly important.

When it comes to achieving the goal of precise image-based 3D modelling, the interior and exterior orientation characteristics of the image are absolutely necessary. For 3D modelling at close range, the relative orientation is sufficient to achieve the desired results. The accuracy of the 3D geometric model that is generated, on the other hand, is directly proportional to the accuracy of the relative orientation parameters (ROPs). In most cases, the photo-triangulation findings obtained from bundle adjustment are used, but robust imaging geometry with high redundancy, good distribution, and accurate tie-point image coordinates is also necessary.

Furthermore, a well-trained professional operator is required for quality assurance purposes, which will reduce the degree of applicability when huge quantities of items are handled. It has been shown that the availability of automatic, dependable, and precise picture matching tools can improve the efficiency of photo-triangulation and surface modelling. Photo-triangulation algorithms have been developed in the field of computer vision, but they are not yet as reliable and exact as those necessary for photogrammetric 3D modelling. Meanwhile, in other cases, the on-site environment or the poor texture of the object surface may make it impossible to obtain evenly distributed tie-point image coordinate measurements, as described above.