China
Africa
Explore chapters and articles related to this topic
Depth map processing method based on edge information for DIBR
Published in Amir Hussain, Mirjana Ivanovic, Electronics, Communications and Networks IV, 2015
XueRui Hu, Ying Yu, Yan Shi, Bo Wang
The depth map is a 2D image that indicates the distance between a world point and the camera. Once a depth map is obtained, we can synthesize a virtual view by 3D warping. The depth value can be calculated from the disparity value by using (1), where v is the intensity of depth map, d is the disparity value of a pixel, dmax and dmin represent the maximum and minimum disparity respectively. v=255⋅d-dmindmax-dmin
Literature Survey on Recent Methods for 2D to 3D Video Conversion
Published in Ling Guan, Yifeng He, Sun-Yuan Kung, Multimedia Image and Video Processing, 2012
Raymond Phan, Richard Rzeszutek, Dimitrios Androutsos
Jung and Ho [57] estimate the depth maps via a Bayesian learning algorithm. Training data of six different features are used, which are the horizontal line, the vanishing point, vertical lines, boundaries, the complexity of the wavelet coefficients, and the object size. This training data categorizes objects in the 2D image into four different types. According to the type, a relative depth value to each object is assumed, and a simple 3D model is generated. The four different types are: sky, ground, plane, and cubic. Sky and ground types are straightforward, whereas the plane type represents an object that is perpendicular to the optical axis of the camera and has a constant depth value. Things like a human and a tree are examples. Finally, the cubic type is regarded as an object that has different depth values, according to the distance from the vanishing point, and includes a building, a wall, and objects such as those. For the rest of the pixels that were not classified, a Bayesian classifier is formulated to choose the best depth values, given those previously mentioned six features.
Vision improvement system using image processing technique for adverse weather condition of opencast mines
Published in International Journal of Mining, Reclamation and Environment, 2019
Debashis Chatterjee, S. K. Chaulya
In the adverse weather conditions, the air is filled with a large amount of suspended particles that scatters the light manifested in the attenuation and enhancement aspects. As the distance between the driver and the point of observation increases, the space between them also rises, presuming that the scattering of suspended particulate matter in the air is uniform each particle in air scatters light. This results the road environment monitored footage being degraded [26]. Depth map contains information linking to the distance of the surfaces of scene objects from a perspective [12]. The obtained depth map is mixed with the depth map of the road surface and other algorithms evaluated in a previous step (Fog Detection Characterisation and vertical object segmentation). In the developed VIS, distance d is applied only for vertical objects, segmented during the same step.Depth map mixing helps simulate the effect of uniformly dense semi-transparent media within a scene in case of fog, haze or smoke.Simulating shallow depths of a road image taken from a dashboard mounted camera.
Dense 3D surface reconstruction of large-scale streetscape from vehicle-borne imagery and LiDAR
Published in International Journal of Digital Earth, 2021
Xiaohu Lin, Bisheng Yang, Fuhong Wang, Jianping Li, Xiqi Wang
When performing incremental dense 3D reconstruction of large-scale streetscapes, on the one hand, key-frame selection, taking a value every few pixels and the voxel grid filtering are introduced to reduce the amount of data and improve the efficiency of storage and display. On the other hand, because of occlusion, illumination change, camera jitter and other objective conditions, there will always be a great quantity of noise in the estimated depth map, which not only has a negative impact on reconstruction accuracy, but also increases the amount of data. In order to get high quality reconstruction results, edge & sky filtering (Kuhn, Lin, and Erdler 2019) and statistical outlier filtering (SOF) are used to remove noise points during dense 3D surface reconstructions.
Design and Development of a Low-Cost Vision-Based 6 DoF Assistive Feeding Robot for the Aged and Specially-Abled People
Published in IETE Journal of Research, 2023
Priyam Parikh, Reena Trivedi, Jatin Dave, Keyur Joshi, Dipak Adhyaru
The depth camera has two vision sensors [50], which are spaced at a small distance, 50 mm in this case [51]. These two vision sensors are identical and configured with identical settings. They have global shutter for supporting high speed applications and a focal length of 1.93 meters. The wide infrared projector available in the camera covers the 99 ± 3 degrees diagonal field of projection improves the ability to determine depth by projecting an infrared pattern on the object to increase texture on low texture objects. The color sensor [52] in the camera provides information on color as well as texture.it has rolling shutter and 1.88 mm of focal length. The diagonal field of view for this sensor is 77 degrees. The depth camera measures depth accurately in the range of 0.3 m to 3 m at a resolution of 1280 × 720. Depth measurements outside this range are not reliable. The depth camera uses light and estimates the depth map using stereo vision principle. The two vision sensors discussed early takes two images in total; one image by each vision sensor. The depth camera that uses stereo principle is known as stereo camera. The stereo camera compares the two images. As the distance between the two vision sensors is known in advance, comparing both images provide depth estimation. The stereo depth camera works in a similar way as we use two eyes for depth perception. In this analogy, our brain calculates the difference between two images by our two eyes. The closer objects appears to move significantly from one vision sensor to the other while the far objects would appear to move little relatively. More information on stereo vision can be obtained from [53]. The infrared projector helps in low light conditions where the depth camera can still perceive depth details as compared with general light. The depth estimation using stereo vision works on the principle of triangulation which is explained in Figure 8. The baseline represents fixed distance between two vision sensors. The depth information for the Cartesian coordinate Z can be obtained from following equation where B is baseline; are coordinates of pixels and is the focal distance as estimated by a calibration process. Figures 10 and 11 shows the plots of estimated depth v/s original depth and original depth v/s error of estimation respectively. The plot highlights that as distance increases, the depth estimation error reduces. However, the depth measurement must lie between the acceptable range for a given depth camera. The results won’t be reliable when the object is near to the depth camera as it provides high error. Such range varies between camera models/manufacturers. The RGB response of the camera at 50 and 100 cm can be seen in Figures 9[(a),(b)]. The complete setup of the vision feeding robotic system with the user can be found in Figure 9(c). Relation between original and estimated depth is shown in Table 9.