Marco Gavelli

Sensor aided H.264 Video Encoder for UAV applications

This paper presents a new low-complexity H.264 encoder for Unmanned Aerial Vehicles (UAV) applications. Standard video coding systems currently employed in UAV applications do not rely on some peculiarities in terms of scene 3D model and correlation among successive frames. In particular, the observed scene is static, i.e. the camera movement is dominant, and it can often be well approximated with a plane. Moreover, camera position and orientation can be obtained from the navigation system. Therefore, correspondent points on two video frames are linked by a simple homography. The encoder employs a new motion estimation scheme which make use of the global motion information provided by the onboard navigation system. The homography is used in order to initialize the block matching algorithm allowing a more robust motion estimation and a smaller search window, and hence reducing the complexity. Experimental results show that the proposed scheme ouperforms standard H.264 in terms of PSNR and throughput. The results are relevant in low frame rate video coding, which is a typical scenario in UAV behind line-of-sight (BLOS) missions. Experiments open new drections in developing new sensor aided video coding standards.

A drone-based image processing system for car detection in a smart transport infrastructure

In this paper we present a car detection system prototyped within an experimental project. It analyzes video streams recorded by drones flying over an urban environment. The intended final goal is the automatic provision of helpful information, such as the available parking spaces and the level of congestion of the streets. The system has been tested both in a desktop PC and on an embedded system. The experimental results show a significant accuracy and prove the feasibility of novel on-board services.

A sensor aided H.264 encoder tested on aerial imagery for SFM

Standard video coding systems currently employed in UAV (Unmanned Aerial Vehicle) and aerial drone applications do not rely on some peculiarities in terms of scene 3D model and correlation among successive frames. In particular, the observed scene is static, i.e. the camera movement is dominant, and it can often be well approximated with a plane. Moreover, camera position and orientation can be obtained from the navigation system. Therefore, correspondent points on two video frames are linked by a simple homography. This paper presents novel results obtained by a low-complexity sensor aided H.264 encoder, recently developed at CIRA and yet tested on simulated data. The proposed encoder employs a new motion estimation scheme which make use of the global motion information provided by the onboard navigation system. The homography is used in order to initialize the block matching algorithm allowing a more robust motion estimation and a smaller search window, and hence reducing the complexity. The tests are made coding real aerial imagery, captured to be used for 3D scene reconstruction. The images are acquired by an high resolution camera mounted on a small drone, flying at low altitude.

An H.264 Sensor Aided Encoder for Aerial Video Sequences with In-the-Loop Metadata Enhancement

Unmanned Aerial Vehicles UAVs are often employed to capture high resolution images in order to perform image mosaicking and/or 3D reconstruction. Images are usually stored on-board or sent to the ground using still image or video data compression. Still image encoders are preferred when low frame rates are involved, because video coding systems are based on motion estimation and compensation algorithms which fail when the motion vectors are significantly long. The latter is the case of low frame rate videos, in which the overlapping between subsequent frames is very small. In this scenario, UAVs attitude and position metadata from the Inertial Navigation System INS can be employed to estimate global motion parameters without video analysis. However, a low complexity analysis can refine the motion field estimated using only the metadata. In this work, we propose to use this refinement step in order to improve the position and attitude estimation produced by the navigation system with the aim of maximizing the encoder performance. Experiments on both simulated and real world video sequences confirm the effectiveness of the proposed approach.

The webinos Architecture: A Developer’s Point of View

This work describes the architecture proposed by the webinos EU project, which aims at developing software components for the future Internet, in the form of Web Runtime Extensions. It discusses the webinos architecture from a developer’s point of view, presenting an overview of its main advantages, such as context-awareness capabilities and distributed APIs in an intrinsic secure environment. It also shows how these features can in practice prove beneficial to the development of ubiquitous and secure web applications based on standard technologies like HTML, CSS and JavaScript.

Demonstrating the new compact descriptors for visual search (CDVS) standard for image retrieval on mobile devices

The MPEG CDVS (Compact Descriptors for Visual Search) standard promises to enable effective, bandwidth-efficient, image matching and retrieval. In this paper, we describe the issues related to the implementation of such technology in an app for smartphones and demonstrate its application to the problem of guiding a tourist through a urban photo safari. To the best of our knowledge, this work is also the first to provide preliminary figures of CDVS performance on a mobile device.

H.264 Sensor Aided Video Encoder for UAV BLOS Missions

This paper presents a new low-complexity H.264 encoder, based on x264 implementation, for Unmanned Aerial Vehicles (UAV) applications. The encoder employs a new motion estimation scheme which make use of the global motion information provided by the onboard navigation system. The results are relevant in low frame rate video coding, which is a typical scenario in UAV behind line-of-sight (BLOS) missions.

A multimodal user interface using the webinos platform to connect a smart input device to the Web of Things

We propose an innovative overlay architecture among heterogeneous devices that works on different operating systems and an application based on standard web technologies that leverages on such architecture to perform a broad range of functions using audio commands. To achieve this, a secure web runtime platform allows connecting different devices such as an Android mobile device, a single board computer and a Web of Things (WoT) sensor/actuator in a secure and independent manner. Methods and functions have been created in order to capture the audio from the microphone on personal computers and on mobile phones, and to safely transfer files to a single board computer, in which a speech recognition engine is embedded. Therefore, by recording the audio command from one of these devices and performing voice command identification using the speech engine, it is possible to perform different actions previously defined by the user on one of his devices. This work demonstrates how open source platforms can interconnect and operate with proprietary architecture in a complementary and secure manner.

Image-processing assisted characterization of spray injection systems

The objective of this work is to investigate the spray characteristics of a fuel injection nozzle. The analysis is performed by means of a framework which exploits different image processing techniques to provide spray-related data to the operator. Innovative metrics are introduced to increase the accuracy and efficiency of the scheme. Experimental results show that it is possible to automatically get useful information about the spray distribution, asymmetries and key properties together with the capability to measure significant angles and other information to detect anomalies in the injection system.