Comparisons of the three-dimensional model reconstructed using MicMac, PIX4D mapper and Photoscan Pro

Three-dimensional models derived from digital survey techniques are growing and focusing on many areas of application. The three-dimensional model has high geometric accuracy with rich spatial and spectral information. In this study, we will utilize three software for three-dimensional models reconstruction, they are commercial software packages and open source tools. Their advantages and disadvantages will be investigated. The suggested commercial suites are PhotoscanPro and Pix4DMapper, whereas the open source one is the MicMac tools developed by IGN France. The images are taken by a Sony Alpha 6000 digital camera with 20 mm focal length. In general, there are four steps to reconstruction a 3D model from 2D digital images. The first step is the initial processing for photo triangulation with camera self-calibration. After initial processing we can obtain the exterior and interior orientation parameters for all images. We setup a scale bar during image acquisition. Thus, after dense image matching, i.e. the second step, we can obtain dense point cloud with real dimension. For accuracy evaluation, we utilize Riegl VZ400 terrestrial laser scanner to collect reference point cloud. Then, we can compare their 3D discrepancy. The third and fourth steps are mesh and texture generation, respectively. The performance comparisons will consider geometric accuracy, completeness and point density. After experiments, all three software provide satisfying results and we conclude that PhotoScanPro and Pix4DMapper are more straightforward to use with user friendly interface. While the MicMac is recommended for experienced users, particularly it support push-broom sensors, such as high-resolution satellite image.