This paper addresses the problem of using live and pre-recorded video sources in 3D reconstruction systems. Typical photogrammetry packages use images generated from digital photography cameras such as DSLRs. 3D recon...
详细信息
ISBN:
(纸本)9781509020508
This paper addresses the problem of using live and pre-recorded video sources in 3D reconstruction systems. Typical photogrammetry packages use images generated from digital photography cameras such as DSLRs. 3D reconstruction algorithms require the intrinsic information to be stored in the EXIF header of a digital photograph. Live video systems don't store this information and more importantly don't produce a concise sequence of distinct frames. This paper presents an approach to using both live and pre-recorded video sources with 3D Structure from Motion systems through the implementation of a pre-processor. This approach filters the video feed with the aid of the Laplacian operator and feature-based methods to create an optimal sequence for both on-line and off-line 3D reconstruction systems.
The current development in imageprocessing towards biometrics systems has opened much research on realtime applications. The deep learning algorithms are added many expectations to the researchers. The main challenge...
详细信息
The detection of targets in military and security applications involves the usage of sensor systems which consist of a variety of sensors such as seismic, acoustic, magnetic and image ones as well. In order to extract...
详细信息
ISBN:
(纸本)9788080405298
The detection of targets in military and security applications involves the usage of sensor systems which consist of a variety of sensors such as seismic, acoustic, magnetic and image ones as well. In order to extract signal features, which characterize particular targets, using of appropriate signal processingalgorithms is required. Seismic signals can be considered as nonstationary and nonlinear signals especially in near-field seismic zone. Most of the signal processingalgorithms assumed that signals are linear and stationary. However, in many cases this assumption is not valid, especially in case of seismic signals generated by moving vehicles, walking persons or gunfire activity. There are several methods which can be used for seismic signal processing, like short-time Fourier transform (STFT), Wavelet transform (WT) and Wigner-Ville distribution (WVD). The paper presents the concept of the seismic sensor system based on Micro-Electro-Mechanical-System (MEMS) sensor SF1500S.A dedicated to vehicle detection. The main part of the paper deals with application of the Hilbert-Huang transform (HHT) to seismic signal processing in time and time-frequency domain. In conclusion, the outcomes of experiments provide comparison of HHT and STFT efficiency in terms of seismic features description of moving vehicle.
image registration is one of the critical techniques in remote sensing imageprocessing. High resolution satellite images are generally obtained at widely spaced intervals. Typically feature-based registration algorit...
详细信息
ISBN:
(纸本)9781479919482
image registration is one of the critical techniques in remote sensing imageprocessing. High resolution satellite images are generally obtained at widely spaced intervals. Typically feature-based registration algorithms may be not proper for those images because of its heavy computation in feature extraction and feature matching process. Additionally, it is unnecessary to use the numerous feature points when calculating image transformation model. In this paper, a registration method based on control points is proposed, which has an excellent performance in large size remote sensing images and the matching process works very fast with the use of a control network. Massive experiments show that the algorithm is effective and efficient and has a strong adaptation to translation, rotation and scale change between images.
Vision based runway identification using 'marked or unmarked terrain' image sequences captured from a fixed wing unmanned aerial vehicle through onboard stereovision sensor is presented in this paper. An innov...
详细信息
Vision based runway identification using 'marked or unmarked terrain' image sequences captured from a fixed wing unmanned aerial vehicle through onboard stereovision sensor is presented in this paper. An innovative convolutional neural netwok (CNN) based YOLO-V8 object detection algorithm is used to detect the runway during approach segment of UAV. This deep learning algorithm detects the region of interest in real time and in a computationally efficient manner. The captured unknown road segment or runway image frames are processed and examined for width, length, level and smoothness aspects to qualify as a suitable runway for UAV landings. Also, it is ensured that there are no obstacles, patches or holes on the detected road or runway. Runway start and end threshold lines and regions, touchdown point and runway edge lines are considered as the region of interest. imageprocessingalgorithms are applied on the captured runway or road images to detect strong features in the region of interest. Feature detector based imageprocessing algorithm with stereo vision constraint is used to establish the relation between unmanned aerial vehicle's center of gravity and detected runway feature points imageprocessingalgorithms like hough line detection, RANSAC, Oriented FAST and Rotated BRIEF (ORB), median filters, morphological methods are applied to extract terrain features. Based on the detected runway orientation and position with respect to UAV position. An automatic landing manoeuvre is performed by UAV autopilot to land the UAV on intended touchdown point on runway computed through detected feature points.
Automotive companies are investing a relevant amount of resources for designing autonomous driving systems, driver assistance technologies, and systems for assessing the driver's attention. In this context, two im...
详细信息
ISBN:
(纸本)9781728157306
Automotive companies are investing a relevant amount of resources for designing autonomous driving systems, driver assistance technologies, and systems for assessing the driver's attention. In this context, two important applications consist in processingimages of the surrounding environment to respectively separate the different objects in the scene (semantic segmentation) and to estimate their distances. In both applications, methods based on Deep Learning (DL) and Convolutional Neural Networks (CNN) are being increasingly used, considering LiDAR scans or RGB images. However, LiDAR scanners require dedicated sensors, high costs, and post-processingalgorithms to estimate a dense depth map or a three-dimensional representation of the surrounding environment. Moreover, current methods in the literature based on RGB images do not consider the combination of semantic segmentation and depth estimation for assessing the distances of specific objects in the scene. In this paper, we propose the first method in the literature able to estimate the distances of pedestrians/cyclists from the vehicle by using only an RGB image and CNNs, without the need for any LiDAR scanner or any device designed for the three-dimensional reconstruction of the scene. We evaluated our approach on a public dataset of RGB images captured in an automotive scenario, with results confirming the feasibility of the proposed method.
Modern cooperative Advanced Driver Assistance systems (ADASs) require efficient algorithms and methods for the real-time processing of all sensor data. In particular, systems for combination of different sensors are g...
详细信息
ISBN:
(纸本)9781479916344
Modern cooperative Advanced Driver Assistance systems (ADASs) require efficient algorithms and methods for the real-time processing of all sensor data. In particular, systems for combination of different sensors are gaining increasing importance. This paper proposes an early fusion technique to combine the advantages of two or more vision sensors. It creates a single composite image that will be more comprehensive for further computer vision tasks, e.g. Object Detection. After image registration, the presented pixel-based fusion framework transforms the registered sensor images into a common representational format by a multiscale decomposition. A denoising and a multiscale edge detection are applied on the transformed data. Only data with a high Activity Level are considered for the fusion process that based on a probabilistic approach. Finally, the fused image can be the input for a subsequent feature extraction task. The proposed fusion technique is examined on a Pedestrian Detection System based on an infrared and a visible light camera.
The development of portable devices such as digital cameras and mobile phones led to higher amount of research being dedicated to image and video processingsystems. In particular, video processing coding technique su...
详细信息
ISBN:
(数字)9781538692790
ISBN:
(纸本)9781538692790
The development of portable devices such as digital cameras and mobile phones led to higher amount of research being dedicated to image and video processingsystems. In particular, video processing coding technique such as High Efficiency Video Coding (HEVC) requires low power consumption for the multimedia applications, this leads to extensive development in low area and high speed algorithms. The log time Discrete Cosine Transform (DCT) is increasingly employed for compression standards, since it has remarkable energy compaction properties. Pipelined and MAC based DCT have been implemented in digital hardware to offer better compression at very low circuit complexity. In this paper, a Time division multiplexing (TDM) based DCT architecture is implemented using the concept of resource sharing principle which possess low computational complexity and high speed. This hardware implementation results in reduced area and increased speed compared to the conventional DCT architectures implemented in Field Programmable Gate Array (FPGA) devices.
In this paper, a new method to predict the classification error in the Gaussian ML classifier is proposed. The Gaussian ML classifier is one of the most widely classifiers in pattern classification and remote sensing ...
详细信息
ISBN:
(纸本)0819425818
In this paper, a new method to predict the classification error in the Gaussian ML classifier is proposed. The Gaussian ML classifier is one of the most widely classifiers in pattern classification and remote sensing because of its speed and performance. Several methods have been proposed to estimate error of the Gaussian ML classifier. In particular, the Bhattacharyya distance gives theoretical upper and lower bounds of the classification error. However, in many cases, the bounds are not tight enough to be useful. In this paper, we propose a different approach to predict error of the Gaussian ML classifier using the Bhattacharyya distance. We generate two classes with normal distribution and calculate the Bhattacharyya distance and the classification accuracy. The class statistics used to generate data are obtained from real remotely sensed data. We repeat the experiment about 100 million times with different class statistics and try to find the relationship between the classification error and the Bhattacharyya distance empirically. The range of the dimension of the generated data is from 1 to 17. From the experiments, we are able to obtain a formula that gives a much better error estimation of the Gaussian ML classifier. Apparently, it is possible to predict the classification error within 1-2% margin.
Motivation: Sensitivity analysis and parameter tuning are important processes in large-scale image analysis. They are very costly because the image analysis workflows are required to be executed several times to syste...
详细信息
Motivation: Sensitivity analysis and parameter tuning are important processes in large-scale image analysis. They are very costly because the image analysis workflows are required to be executed several times to systematically correlate output variations with parameter changes or to tune parameters. An integrated solution with minimum user interaction that uses effective methodologies and high performance computing is required to scale these studies to large imaging datasets and expensive analysis workflows. Results: The experiments with two segmentation workflows show that the proposed approach can (i) quickly identify and prune parameters that are non-influential;(ii) search a small fraction (about 100 points) of the parameter search space with billions to trillions of points and improve the quality of segmentation results (Dice and Jaccard metrics) by as much as 1.42 x compared to the results from the default parameters;(iii) attain good scalability on a high performance cluster with several effective optimizations. Conclusions: Our work demonstrates the feasibility of performing sensitivity analyses, parameter studies and auto-tuning with large datasets. The proposed framework can enable the quantification of error estimations and output variations in image segmentation pipelines.
暂无评论