Search

Neural radiance fields (NeRF) have become an effective method for encoding scenes into neural representations, allowing for the synthesis of photorealistic views of unseen views from given input images. However, the applicability of traditional NeRF is significantly limited by its assumption that images are captured for object-centric scenes with a pinhole camera. Expanding these boundaries, we focus on driving scenarios using a fisheye camera, which offers the advantage of capturing visual information from a wide field of view. To address the challenges due to the unbounded and distorted characteristics of fisheye images, we propose an edge-aware integration loss function. This approach leverages sparse LiDAR projections and dense depth maps estimated from a learning-based depth model. The proposed algorithm assigns larger weights to neighboring points that have depth values similar to the sensor data. Experiments were conducted on the KITTI-360 and JBNU-Depth360 datasets, which are public and real-world datasets of driving scenarios using fisheye cameras. Experimental results demonstrated that the proposed method is effective in synthesizing novel view images, outperforming existing approaches. [ABSTRACT FROM AUTHOR]

Ground target detection and positioning systems based on lightweight unmanned aerial vehicles (UAVs) are increasing in value for aerial reconnaissance and surveillance. However, the current method for estimating the target's position is limited by the field of view angle, rendering it challenging to fulfill the demands of a real-time omnidirectional reconnaissance operation. To address this issue, we propose an Omnidirectional Optimal Real-Time Ground Target Position Estimation System (Omni-OTPE) that utilizes a fisheye camera and LiDAR sensors. The object of interest is first identified in the fisheye image, and then, the image-based target position is obtained by solving using the fisheye projection model and the target center extraction algorithm based on the detected edge information. Next, the LiDAR's real-time point cloud data are filtered based on position–direction constraints using the image-based target position information. This step allows for the determination of point cloud clusters that are relevant to the characterization of the target's position information. Finally, the target positions obtained from the two methods are fused using an optimal Kalman fuser to obtain the optimal target position information. In order to evaluate the positioning accuracy, we designed a hardware and software setup, mounted on a lightweight UAV, and tested it in a real scenario. The experimental results validate that our method exhibits significant advantages over traditional methods and achieves a real-time high-performance ground target position estimation function. [ABSTRACT FROM AUTHOR]

Monocular depth estimation is a task aimed at predicting pixel-level distances from a single RGB image. This task holds significance in various applications including autonomous driving and robotics. In particular, the recognition of surrounding environments is important to avoid collisions during autonomous parking. Fisheye cameras are adequate to acquire visual information from a wide field of view, reducing blind spots and preventing potential collisions. While there have been increasing demands for fisheye cameras in visual-recognition systems, existing research on depth estimation has primarily focused on pinhole camera images. Moreover, depth estimation from fisheye images poses additional challenges due to strong distortion and the lack of public datasets. In this work, we propose a novel underground parking lot dataset called JBNU-Depth360, which consists of fisheye camera images and their corresponding LiDAR projections. Our proposed dataset was composed of 4221 pairs of fisheye images and their corresponding LiDAR point clouds, which were obtained from six driving sequences. Furthermore, we employed a knowledge-distillation technique to improve the performance of the state-of-the-art depth-estimation models. The teacher–student learning framework allows the neural network to leverage the information in dense depth predictions and sparse LiDAR projections. Experiments were conducted on the KITTI-360 and JBNU-Depth360 datasets for analyzing the performance of existing depth-estimation models on fisheye camera images. By utilizing the self-distillation technique, the AbsRel and SILog error metrics were reduced by 1.81% and 1.55% on the JBNU-Depth360 dataset. The experimental results demonstrated that the self-distillation technique is beneficial to improve the performance of depth-estimation models. [ABSTRACT FROM AUTHOR]

The fisheye camera, with its large viewing angle, can acquire more spatial information in one shot and is widely used in many fields. However, a fisheye image contains large distortion, resulting in that many scholars have investigated its accuracy of orthorectification, i.e., generation of digital orthophoto map (DOM). This paper presents an orthorectification method, which first determines the transformation relationship between the fisheye image points and the perspective projection points according to the equidistant projection model, i.e., determines the spherical distortion of the fisheye image; then introduces the transformation relationship and the fisheye camera distortion model into the collinearity equation to derive the fisheye image orthorectification model. To verify the proposed method, high accuracy of the fisheye camera 3D calibration field is established to obtain the interior and exterior orientation parameters (IOPs/EOPs) and distortion parameters of the fisheye lens. Three experiments are used to verify the proposed orthorectification method. The root mean square errors (RMSEs) of the three DOMs are averagely 0.003 m, 0.29 m, and 0.61 m, respectively. The experimental results demonstrate that the proposed method is correct and effective. [ABSTRACT FROM AUTHOR]

Solar radiation is one of the most significant environmental factors that regulates the rate of photosynthesis, and consequently, growth. Light intensity in the forest can vary both spatially and temporally, so precise assessment of canopy and potential solar radiation can significantly influence the success of forest management actions, for example, the establishment of natural regeneration. In this case study, we investigated the possibilities and perspectives of close-range photogrammetric approaches for modeling the amount of potential direct and diffuse solar radiation during the growing seasons (spring-summer), by comparing the performance of low-cost Unmanned Aerial Vehicle (UAV) RGB imagery vs. Hemispherical Photography (HP). Characterization of the solar environment based on hemispherical photography has already been widely used in botany and ecology for a few decades, while the UAV method is relatively new. Also, we compared the importance of several components of potential solar irradiation and their impact on the regeneration of Pinus sylvestris L. For this purpose, a circular fisheye objective was used to obtain hemispherical images to assess sky openness and direct/diffuse photosynthetically active flux density under canopy average for the growing season. Concerning the UAV, a Canopy Height Model (CHM) was constructed based on Structure from Motion (SfM) algorithms using Photoscan professional. Different layers such as potential direct and diffuse radiation, direct duration, etc., were extracted from CHM using ArcGIS 10.3.1 (Esri: California, CA, USA). A zonal statistics tool was used in order to extract the digital data in tree positions and, subsequently, the correlation between potential solar radiation layers and the number of seedlings was evaluated. The results of this study showed that there is a high relation between the two used approaches (HP and UAV) with R2 = 0.74. Finally, potential diffuse solar radiation derived from both methods had the highest significant relation (-8.06% bias) and highest impact in the modeling of pine regeneration. [ABSTRACT FROM AUTHOR]

This paper presents a practical application of a technique that uses a vertical optical flow with a fisheye camera to generate dense point clouds from a single planimetric station. Accurate data can be extracted to enable the measurement of tree trunks or branches. The images that are collected with this technique can be oriented in photogrammetric software (using fisheye models) and used to generate dense point clouds, provided that some constraints on the camera positions are adopted. A set of images was captured in a forest plot in the experiments. Weighted geometric constraints were imposed in the photogrammetric software to calculate the image orientation, perform dense image matching, and accurately generate a 3D point cloud. The tree trunks in the scenes were reconstructed and mapped in a local reference system. The accuracy assessment was based on differences between measured and estimated trunk diameters at different heights. Trunk sections from an image-based point cloud were also compared to the corresponding sections that were extracted from a dense terrestrial laser scanning (TLS) point cloud. Cylindrical fitting of the trunk sections allowed the assessment of the accuracies of the trunk geometric shapes in both clouds. The average difference between the cylinders that were fitted to the photogrammetric cloud and those to the TLS cloud was less than 1 cm, which indicates the potential of the proposed technique. The point densities that were obtained with vertical optical scanning were 1/3 less than those that were obtained with TLS. However, the point density can be improved by using higher resolution cameras. [ABSTRACT FROM AUTHOR]

Fisheye images with a far larger Field of View (FOV) have severe radial distortion, with the result that the associated image feature matching process cannot achieve the best performance if the traditional feature descriptors are used. To address this challenge, this paper reports a novel distorted Binary Robust Independent Elementary Feature (BRIEF) descriptor for fisheye images based on a spherical perspective model. Firstly, the 3D gray centroid of feature points is designed, and the position and direction of the feature points on the spherical image are described by a constructed feature point attitude matrix. Then, based on the attitude matrix of feature points, the coordinate mapping relationship between the BRIEF descriptor template and the fisheye image is established to realize the computation associated with the distorted BRIEF descriptor. Four experiments are provided to test and verify the invariance and matching performance of the proposed descriptor for a fisheye image. The experimental results show that the proposed descriptor works well for distortion invariance and can significantly improve the matching performance in fisheye images. [ABSTRACT FROM AUTHOR]

The development of information technology has added many conveniences to our lives. On the other hand, however, we have to deal with various kinds of information, which can be a difficult task for elderly people or those who are not familiar with information devices. A technology to recognize each person's activity and providing appropriate support based on that activity could be useful for such people. In this paper, we propose a novel fine-grained activity recognition method for user support systems that focuses on identifying the text at which a user is gazing, based on the idea that the content of the text is related to the activity of the user. It is necessary to keep in mind that the meaning of the text depends on its location. To tackle this problem, we propose the simultaneous use of a wearable device and fixed camera. To obtain the global location of the text, we perform image matching using the local features of the images obtained by these two devices. Then, we generate a feature vector based on this information and the content of the text. To show the effectiveness of the proposed approach, we performed activity recognition experiments with six subjects in a laboratory environment. [ABSTRACT FROM AUTHOR]