Multi-view Stereo (MVS) aims to estimate depth and reconstruct 3D point clouds from a series of overlapping images. Recent learning-based MVS frameworks overlook the geometric information embedded in features and correlations, leading to weak cost matching. In this paper, we propose ICG-MVSNet, which explicitly integrates intra-view and cross-view relationships for depth estimation. Specifically, we develop an intra-view feature fusion module that leverages the feature coordinate correlations within a single image to enhance robust cost matching. Additionally, we introduce a lightweight cross-view aggregation module that efficiently utilizes the contextual information from volume correlations to guide regularization. Our method is evaluated on the DTU dataset and Tanks and Temples benchmark, consistently achieving competitive performance against state-of-the-art works, while requiring lower computational resources.

Reliable self-localization is a foundational skill for many intelligent mobile platforms. This paper explores the use of event cameras for motion tracking thereby providing a solution with inherent robustness under difficult dynamics and illumination. In order to circumvent the challenge of event camera-based mapping, the solution is framed in a cross-modal way. It tracks a map representation that comes directly from frame-based cameras. Specifically, the proposed method operates on top of gaussian splatting, a state-of-the-art representation that permits highly efficient and realistic novel view synthesis. The key of our approach consists of a novel pose parametrization that uses a reference pose plus first order dynamics for local differential image rendering. The latter is then compared against images of integrated events in a staggered coarse-to-fine optimization scheme. As demonstrated by our results, the realistic view rendering ability of gaussian splatting leads to stable and accurate tracking across a variety of both publicly available and newly recorded data sequences.

Event cameras are an interesting visual exteroceptive sensor that reacts to brightness changes rather than integrating absolute image intensities. Owing to this design, the sensor exhibits strong performance in situations of challenging dynamics and illumination conditions. While event-based simultaneous tracking and mapping remains a challenging problem, a number of recent works have pointed out the sensor’s suitability for prior map-based tracking. By making use of cross-modal registration paradigms, the camera’s ego-motion can be tracked across a large spectrum of illumination and dynamics conditions on top of accurate maps that have been created a priori by more traditional sensors. The present paper follows up on a recently introduced event-based geometric semi-dense tracking paradigm, and proposes the addition of inertial signals in order to robustify the estimation. More specifically, the added signals provide strong cues for pose initialization as well as regularization during windowed, multi-frame tracking. As a result, the proposed framework achieves increased performance under challenging illumination conditions as well as a reduction of the rate at which intermediate event representations need to be registered in order to maintain stable tracking across highly dynamic sequences. Our evaluation focuses on a diverse set of real world sequences and comprises a comparison of our proposed method against a purely event-based alternative running at different rates.

Pose graph relaxation has become an indispensable addition to SLAM enabling efficient global registration of sensor reference frames under the objective of satisfying pair-wise relative transformation constraints. The latter may be given by incremental motion estimation or global place recognition. While the latter case enables loop closures and drift compensation, care has to be taken in the monocular case in which local estimates of structure and displacements can differ from reality not just in terms of noise, but also in terms of a scale factor. Owing to the accumulation of scale propagation errors, this scale factor is drifting over time, hence scale-drift aware pose graph relaxation has been introduced. We extend this idea to cases in which the relative scale between subsequent sensor frames is unknown, a situation that can easily occur if monocular SLAM enters re-initialization and no reliable overlap between successive local maps can be identified. The approach is realized by a hybrid pose graph formulation that combines the regular similarity consistency terms with novel, scale-blind constraints. We apply the technique to the practically relevant case of small indoor service robots capable of effectuating purely rotational displacements, a condition that can easily cause tracking failures. We demonstrate that globally consistent trajectories can be recovered even if multiple re-initializations occur along the loop, and present an in-depth study of success and failure cases.

In this study, we propose a novel CLAHE-based nighttime image contrast enhancement approach for vehicle detection under nighttime conditions, which improves the contrast of low-quality nighttime images while preventing over-enhancement by employing the image dehazing technique. To implement and evaluate our proposed contrast enhancement method on nighttime images, we consider a scenario of using a camera-based Internet of Things (IoT)-edge computing device for traffic and road surveillance. Edge-computing and IoT technology enable significant amounts of novel studies to advance traffic system monitoring, sensing, control, and management. Considering multiple metrics of image enhancement quality, the proposed nighttime image contrast enhancement method outperforms some existing well-performing CLAHE-based methods. To provide accurate vehicle detection under nighttime conditions and different challenges, including vehicle overlapping, low-light conditions, camera vibrations, and image distortion, must be addressed. For this purpose, a deep neural network based on YOLOv5 architecture has been designed and trained using our custom-labeled dataset. The developed neural network is proven to be effective in the detection of different vehicles under low-light ambient conditions using video captured from a stationary camera. Experiments on our dataset show that the proposed contrast enhancement method greatly improves the detection performance of the trained YOLOv5 model under low-environment-light conditions compared with the model trained using unenhanced images. The model trained with enhanced images can provide an improvement of 5.7% on F1 score, 6.3% on mAP0.5, and 3.4% on mAP0.5:0.95 under specific conditions.

A significant number of camera-based solutions suffer from different kinds of performance issues, caused by unstable weather conditions e.g., wind, object overlapping due to heavy traffic flow, and lack of adjustment for a certain location where the camera is installed and fixed. In this paper, we developed an effective approach to traffic flow monitoring under daytime conditions by applying machine learning and computer vision techniques to extract motion traffic data parameters from the videos captured by the static surveillance camera installed and fixed at the intersection. We address the issue of video camera vibration and jittering by applying image-matching algorithms. In order to calculate the vehicle speed, we utilize the projective transformation to compute the real object distance from an image taken by a single camera installed at the road intersection. We adapt and employ state-of-the-art object detectors, transfer learning, and analytical computer vision methods to detect and track vehicles and measure moving speed and vehicle volume from a predefined detection area extracted from monocular videos. The developed approach is proven to be effective in estimating vehicle speed and vehicle volume using video sequences captured from a stationary camera. Experimental results obtained during the study prove the efficiency of the proposed method under strong vibration conditions. The proposed framework achieved promising tracking performance on custom-labeled test video clips with a small absolute error of 3.97% for traffic flow average speed estimation. An additional test on a video from BrnoCompSpeed Dataset shows that the proposed method can reduce the average speed estimation error by 51.8%.