Department of Electrical and Computer Engineering, Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, NL, Canada
Mr. Ali Ebrahimi
Department of Electrical and Computer Engineering, Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, NL, Canada
Removing bounding surfaces such as walls, windows, curtains, and floor (i.e., super-surfaces) from a point cloud is a common task in a wide variety of computer vision applications (e.g., object recognition and human tracking). Popular plane segmentation methods such as Random Sample Consensus (RANSAC), are widely used to segment and remove surfaces from a point cloud. However, these estimators easily result in the incorrect association of foreground points to background bounding surfaces because of the stochasticity of randomly sampling, and the limited scene-specific knowledge used by these approaches. Additionally, identical approaches are generally used to detect bounding surfaces and surfaces that belong to foreground objects. Detecting and removing bounding surfaces in challenging (i.e., cluttered and dynamic) real-world scene can easily result in the erroneous removal of points belonging to desired foreground objects such as human bodies. To address these challenges, we introduce a novel super-surface removal technique for 3D complex indoor environments. Our method was developed to work with unorganized data captured from commercial depth sensors and supports varied sensor perspectives. We begin with preprocessing steps and dividing the input point cloud into four overlapped local regions. Then, we apply an iterative surface removal approach to all four regions to segment and remove the bounding surfaces. We evaluate the performance of our proposed method in terms of four conventional metrics: specificity, precision, recall, and F1 score, on three generated datasets representing different indoor environments. Our experimental results demonstrate that our proposed method is a robust super-surface removal and size reduction approach for complex 3D indoor environments while scoring the four evaluation metrics between 90% and 99%.
21 January 2015 in Neural Computing and Applications
In today's digital era, because of the advent of image editing and processing tools it is conceivable to alter digital images very easily without leaving any obvious tampering traces. Thus, image forgery has become very easy and authenticity of digital images has been severely threatened, resulting in a need for reliable and robust image forensic techniques. Image painting is one of the commonly used methods to conceal unwanted portions of an image. In this paper, a passive method for detecting painted areas on widely used JPEG images is proposed. The block processing during JPEG compression presents horizontal and vertical breaks into images, which is recognized as block artifact grids (BAGs). The detection technique is based on the fact that BAGs usually disarrange after performing painting operations. Experimental results demonstrate the validity and effectiveness of the proposed approach.
The emergence of commercially accessible depth sensors has driven the widespread adoption of 3D data, offering substantial benefits across diverse applications, ranging from human activity recognition to augmented reality. However, indoor environments present significant challenges for 3D computer vision applications, particularly in cluttered and dynamic scenes where background bounding surfaces hinder the detection and analysis of foreground objects. We introduce a novel perspective-independent point cloud simplifier (PiPCS) for complex 3D indoor scenes. PiPCS streamlines 3D computer vision workflows by contextually segmenting and subtracting background bounding surfaces and preserving segmented foreground objects within indoor scenes, effectively reducing the size of indoor point clouds and enhancing 3D indoor scene perception. Methodologically, we use estimated surface normals to intelligently divide an input point cloud into distinct zones, which are then split into multiple distinct parallel clusters. Next, we find the largest plane in each cluster and sort the fitted planes within each zone based on their distance along the zone’s normal vector to identify the bounding surfaces. Finally, we segment the 3D background, simplify the point cloud by employing a voxel-based background subtraction technique, and segment 3D foreground objects via a cluster-based segmentation approach. We evaluated PiPCS on the Stanford S3DIS dataset and our own challenging dataset, achieving average values of 97.08% for specificity and 91.27% for F1 score on the S3DIS dataset and size reductions averaging 74.11% overall. Our experimental and evaluation results demonstrate that PiPCS robustly simplifies and reduces the size of unorganized indoor point clouds.