Low-performance Hardware Research Articles

Real World Object Detection Dataset for Quadcopter Unmanned Aerial Vehicle Detection

Recent years have shown a noticeable rise in the number of incidents with drones, related to both civilian and military installations. While drone neutralization techniques have become increasingly effective, detection most often relies on professional equipment, which is too expensive to be used for all critical nodes and applications. Therefore, there is a need for drone detection systems that could work on low performance hardware. Its critical component consists of an object detection system. In this article, we introduce a new object detection dataset, built entirely to train computer vision based object detection machine learning algorithms for a task of binary object detection to enable automated, industrial camera based detection of multiple drone objects using camera feed. The dataset expands existing multiclass image classification and object detection datasets (ImageNet, MS-COCO, PASCAL VOC, anti-UAV) with a diversified dataset of drone images. In order to maximize the effectiveness of the model, real world footage was utilized, transformed into images and hand-labelled to create a custom set of 56821 images and 55539 bounding boxes. Additionally, semi-automated labelling was proposed, tested and proved to be very useful for object detection applications. The dataset was divided into train and test subsets for further processing and used to generate 603 easily deployable Haar Cascades as well as 819 high performing Deep Neural Networks based models. They were used to test different object detection methods to determine the long term feasibility of a large scale drone detection system utilizing machine learning algorithms. The study has shown that Haar Cascade can be used as the Minimum Viable Product model for mediocre performance but fails to scale up effectively for a larger dataset compared to the Deep Neural Network model.

A Low Power and High Performance Hardware Design for Automatic Epilepsy Seizure Detection

A Low Power and High Performance Hardware Design for Automatic Epilepsy Seizure Detection

Data Mining of Spatio-Temporal Variability of Chlorophyll-a Concentrations in a Portion of the Western Atlantic with Low Performance Hardware

The contemporary scientific literature that deals with the dynamics of marine chlorophyll-a concentration is already customarily employing data mining techniques in small geographic areas or regional samples. However, there is little focus on the issue of missing data related to chlorophyll-a concentration estimated by remote sensors. Intending to provide greater scope to the identification of the spatiotemporal distribution patterns of marine chlorophyll-a concentrations, and to improve the reliability of results, this study presents a data mining approach to cluster similar chlorophyll-a concentration behaviors while implementing an iterative spatiotemporal interpolation technique for missing data inference. Although some dynamic behaviors of said concentrations in specific areas are already known by specialists, systematic studies in large geographical areas are still scarce due to the computational complexity involved. For this reason, this study analyzed 18 years of NASA satellite observations in one portion of the Western Atlantic Ocean, totaling more than 60 million records. Additionally, performance tests were carried out in low-cost computer systems to check the accessibility of the proposal implemented for use in computational structures of different sizes. The approach was able to identify patterns with high spatial resolution, accuracy and reliability, rendered in low-cost computers even with large volumes of data, generating new and consistent patterns of spatiotemporal variability. Thus, it opens up new possibilities for data mining research on a global scale in this field of application.

Harmonic and interharmonic measurements through a compressed sampling approach

The paper deals with the distributed monitoring of harmonic and interharmonic pollution in electrical power delivery systems. In order to measure the disturbances and to identify the sources in a wide grid, a distributed measurement system with hundreds of measurement nodes has to be adopted. With the aim of obtaining the reduction of realization costs, the authors propose an innovative distributed architecture, based on cost-effective nodes, that takes advantage from the compressive sampling strategy. Differently from traditional approach, the network nodes will be only mandated to acquire and to digitize the line voltage directly in a compressed form, whereas a central measurement unit will receive data from all nodes deployed in the grid and will perform the successive signal reconstruction, making possible the use of low-performance hardware to realize them. The assessment of the compliance of the proposed measurement technique with the current power quality standards turns out to be mandatory, thus verifying the absence of artifacts introduced by the adopted compressed sampling approach.

Fast and Dense Depth Map Estimation for Stereovision Low-cost Systems

Depth map extraction or estimation is very important for many computer vision applications. However, this process involves high cost. Hence, it is interesting to find new approaches, which could yield faster implementations, when very accurate results are not necessary. These new implementations can be built up for the stereovision problem, taking advantage of the specific geometrical constraints of this kind of problem. The problem is solved by searching pixel-by-pixel matching in horizontal lines. In this article, we propose a novel approach to this problem, providing measurements of the processing time and accuracy achieved, as well as a qualitative comparison with other methods available in the literature. The results obtained could reach speeds around one order of magnitude faster than those proposed in other classical implementations. Thus, the proposed algorithm can run on cheap and low-performance hardware achieving real-time processing rates.

Evaluation of magnetic nanoparticle samples made from biocompatible ferucarbotran by time-correlation magnetic particle imaging reconstruction method

BackgroundMolecular imaging using magnetic nanoparticles (MNPs)—magnetic particle imaging (MPI)—has attracted interest for the early diagnosis of cancer and cardiovascular disease. However, because a steep local magnetic field distribution is required to obtain a defined image, sophisticated hardware is required. Therefore, it is desirable to realize excellent image quality even with low-performance hardware. In this study, the spatial resolution of MPI was evaluated using an image reconstruction method based on the correlation information of the magnetization signal in a time domain and by applying MNP samples made from biocompatible ferucarbotran that have adjusted particle diameters.MethodsThe magnetization characteristics and particle diameters of four types of MNP samples made from ferucarbotran were evaluated. A numerical analysis based on our proposed method that calculates the image intensity from correlation information between the magnetization signal generated from MNPs and the system function was attempted, and the obtained image quality was compared with that using the prototype in terms of image resolution and image artifacts.ResultsMNP samples obtained by adjusting ferucarbotran showed superior properties to conventional ferucarbotran samples, and numerical analysis showed that the same image quality could be obtained using a gradient magnetic field generator with 0.6 times the performance. However, because image blurring was included theoretically by the proposed method, an algorithm will be required to improve performance.ConclusionsMNP samples obtained by adjusting ferucarbotran showed magnetizing properties superior to conventional ferucarbotran samples, and by using such samples, comparable image quality (spatial resolution) could be obtained with a lower gradient magnetic field intensity.

완전한 3차원을 지원하는 온라인 RPG를 위한 맵 관리 방법

In this paper we suggested a map management method for on line RPG supporting full 3D. In general, it is possible to support full 3D with client engine. Rut the online game server cannot show the expected performance under normal price constrains with current hardware technology, since the amount of data for management of 3D on-line game server increase exponentially with respect to the size of game world. To solve this problem. we introduced the special object which makes it possible for on-line game server with low performance hardware. This suggested method gave concrete from to manage full 3D for server as well as clients.s clients.