![]() ![]() to identify the ones with the highest effect on the resulting depth maps. Both the Intel cameras, are evaluated on all possible combinations of the available post-processing filters such as highDensity (prioritize more depth values), highAccuracy (prioritise accurate depth values), etc. Therefore it strongly depends on good lighting conditions and well-represented texture on the objects. In contrast to the D415, the D405 is only based on stereo vision and has no additional infra-red emitter. ![]() The Intel RealSense™ D405 is the newest among the cameras, and is designed for applications where good accuracy and precision are important in the close-range, such as inspection and high precision picking and placing of small objects. The camera has a rolling shutter which improves the quality of depth measurements on static scenes. The Intel RealSense™ D415 is an active stereoscopic depth camera introduced in 2018, with an infra-red emitter for active depth measurements. Selected technical features of the respective cameras are provided in Table Table1. Furthermore, the influence of several parameters such as resolution, distance from the camera, and camera modes, is systematically evaluated to assess the suitability to different scenes and objects in a close-range setting.įigure 1 shows the three different depth cameras that are compared in this work. The environments comprise static and dynamic scenes, that mimic surgical simulation in MVR. An evaluation is performed, across three different environments containing planar surfaces (Env01), rigid objects of known geometry (Env02), patient-specific silicone mitral valve replica and porcine valves (Env03). The performance of the 3 cameras are compared for use in surgical simulation. In this work, three different commercially available depth sensors are identified that potentially work in the close-range as per manufacturer specifications, namely the Intel RealSense™ D415, the Intel Realsense™ D405 (Intel Corporation, Santa Clara, US), and the Stereolabs ZED-Mini (StereoLabs, San Francisco, US). Particularly in MVR, where surgical training is performed using patient-specific surgical simulators, wet surfaces and reflective materials potentially play a role in a robust 3D reconstruction of the scene. This hinders their adoption in surgical simulation, where typical applications like identifying fine anatomical structures, making quantitative measurements, and reconstructing dynamic scenes require accurate depth sensing in the close-range. However, most off-the-shelf depth cameras function optimally in the range of 1 to 10 m. In particular in minimally invasive mitral valve repair (MVR), a surgery of the heart-valve, depth information facilitates surgical decision making, for example in choosing an appropriate size of ring prosthesis for the valve. Here, RGB-D sensors, which capture color (RGB) and depth (D) images in real time, show great potential in providing such quantitative information without the need for markers or a complex setup. For instance, the endoscopic video assistance used during surgical simulation can be enhanced with a 3D reconstruction of the scene. Moreover, they are capable of providing quantitative feedback that further improves the surgical training process. Surgical simulators in this regard, have the potential to reduce this steep learning curve, by enabling surgical training. However, maneuvering elongated surgical instruments through narrow ports demands a high skill level and dexterity from the surgeon. Minimally invasive surgeries are increasingly prevalent in the recent years, as the surgical ports get smaller, enabling faster recovery times. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |