A Real-time Algorithm for Visual Detection of High-speed Unmanned Surface Vehicle Based on Deep Learning

This paper introduces a novel low-cost solar-powered wearable assistive technology (AT) device, whose aim is to provide continuous, real-time object recognition to ease the finding of the objects for visually impaired (VI) people in daily life. The system consists of three major components: a miniature low-cost camera, a system on module (SoM) computing unit, and an ultrasonic sensor. The first is worn on the user’s eyeglasses and acquires real-time video of the nearby space. The second is worn as a belt and runs deep learning-based methods and spatial algorithms which process the video coming from the camera performing objects’ detection and recognition. The third assists on positioning the objects found in the surrounding space. The developed device provides audible descriptive sentences as feedback to the user involving the objects recognized and their position referenced to the user gaze. After a proper power consumption analysis, a wearable solar harvesting system, integrated with the developed AT device, has been designed and tested to extend the energy autonomy in the different operating modes and scenarios. Experimental results obtained with the developed low-cost AT device have demonstrated an accurate and reliable real-time object identification with an 86% correct recognition rate and 215 ms average time interval (in case of high-speed SoM operating mode) for the image processing. The proposed system is capable of recognizing the 91 objects offered by the Microsoft Common Objects in Context (COCO) dataset plus several custom objects and human faces. In addition, a simple and scalable methodology for using image datasets and training of Convolutional Neural Networks (CNNs) is introduced to add objects to the system and increase its repertory. It is also demonstrated that comprehensive trainings involving 100 images per targeted object achieve 89% recognition rates, while fast trainings with only 12 images achieve acceptable recognition rates of 55%.

Download Full-text

Theoretical and Empirical Analysis of Crime Data

Journal of Web Engineering ◽

10.13052/jwe1540-9589.2016 ◽

2021 ◽

Author(s):

Manisha Mudgal ◽

Deepika Punj ◽

Anuradha Pillai

Keyword(s):

Data Mining ◽

Neural Networks ◽

Artificial Neural Networks ◽

Deep Learning ◽

Real Time ◽

High Speed ◽

Association Mining ◽

Crime Prediction ◽

Methods And Techniques ◽

Speed And Accuracy

Crime is one of the biggest and dominating problems in today’s world and it is not only harmful to the person involved but also to the community and government. Due to escalation in crime frequency, there is a need for a system that can detect and predict crimes. This paper describes the summary of the different methods and techniques used to identify, analyze and predict upcoming and present crimes. This paper shows, how data mining techniques can be used to detect and predict crime using association mining rule, k-means clustering, decision tree, artificial neural networks and deep learning methods are also explained. Most of the researches are currently working on forecasting the occurrence of future crime. There is a need for approaches that can work on real-time crime prediction at high speed and accuracy. In this paper, a model has been proposed that can work on real-time crime prediction by recognizing human actions.

Download Full-text

Deep Learning for Downhole Data Prediction: A Cost-Effective Data Telemetry Through Data Analytics

10.2118/200870-ms ◽

2021 ◽

Author(s):

Bharat Thakur ◽

Robello Samuel

Keyword(s):

Time Series ◽

Deep Learning ◽

Real Time ◽

High Speed ◽

Cost Effective ◽

Time Series Forecasting ◽

Data Prediction ◽

Surface Data ◽

Median Error

Abstract Accurate real-time downhole data collection provides a better understanding of downhole dynamics and formation characteristics, which can improve wellbore placement and increase drilling efficiency by improving the rate of penetration (ROP) and reducing downtime caused by tool failure. High-speed telemetry through wired drill string has enabled real-time data acquisition, but there are significant additional costs associated with the technology. Data-driven techniques using recursive neural networks (RNN) have proven very efficient and accurate in time-series forecasting problems. In this study, we propose deep learning as a cost-effective method to predict downhole data using surface data. Downhole drilling data is a function of surface drilling parameters and downhole conditions. The downhole data acquired using relatively inexpensive methods usually have a considerable lag time depending on the signal travel length. So, the first step in the proposed method is syncing the downhole data and surface data. After the data are synced, they are then fed into an RNN-based long-term short memory (LSTM) network, which learns the relationship between the surface parameters and downhole data. LSTM networks can learn long-term relationships in the data, thus making them ideal for time-series forecasting applications. The trained model is then used to make predictions for downhole data using the given surface data. The median error for the prediction of downhole data using surface data was as low as 3% in this study. The study suggests that the developed model can accurately predict downhole data in real-time. The model is also very robust to the amount of noise or outliers present in the data and can predict downhole conditions 50–60 ft ahead with reasonable accuracy. It was observed that the prediction accuracy varied from well to well and drilling depths. The results demonstrate how deep learning can be cost-effectively employed for downhole data prediction. This paper presents a novel method for using surface data to predict downhole data by employing deep learning. The method can be deployed in real-time to aid in wellbore placement and improve drilling performance.

Download Full-text

Simulation and Detection of Small Drones/Suspicious UAVs in Drone Grid

International Journal for Research in Applied Science and Engineering Technology ◽

10.22214/ijraset.2021.36144 ◽

2021 ◽

Vol 9 (VI) ◽

pp. 5452-5458

Author(s):

Arpit Gupta

Keyword(s):

Real Time ◽

High Speed ◽

Visual Detection ◽

Autonomous Systems ◽

Sliding Window ◽

The Past ◽

Detection And Tracking ◽

Window Method ◽

Sliding Window Method ◽

State Of Art

Today’s technology is evolving towards autonomous systems and the demand in autonomous drones, cars, robots, etc. has increased drastically in the past years. This project presents a solution for autonomous real-time visual detection and tracking of hostile drones by moving cameras equipped on surveillance drones. The algorithm developed in this project, based on state-of-art machine learning and computer vision methods, succeeds at autonomously detecting and tracking a single drone by moving a camera and can run at real-time. The project can be divided into two main parts: the detection and the tracking. The detection is based on the YOLOv3 (You Only Look Once v3) algorithm and a sliding window method. The tracking is based on the GOTURN (Generic Object Tracking Using Regression Networks) algorithm, which allows the tracking of generic objects at high speed. In order to allow autonomous tracking and enhance the accuracy, a combination of GOTURN and tracking by detection using YOLOv3 was developed.

Download Full-text

Smart teaching mode based on particle swarm image recognition and human-computer interaction deep learning

Journal of Intelligent & Fuzzy Systems ◽

10.3233/jifs-189048 ◽

2020 ◽

Vol 39 (4) ◽

pp. 5699-5711

Author(s):

Shirong Long ◽

Xuekong Zhao

Keyword(s):

Feature Extraction ◽

Particle Swarm Optimization ◽

Deep Learning ◽

Real Time ◽

Image Recognition ◽

Particle Swarm ◽

Learning Technology ◽

Search Performance ◽

Swarm Optimization ◽

Teaching Mode

The smart teaching mode overcomes the shortcomings of traditional teaching online and offline, but there are certain deficiencies in the real-time feature extraction of teachers and students. In view of this, this study uses the particle swarm image recognition and deep learning technology to process the intelligent classroom video teaching image and extracts the classroom task features in real time and sends them to the teacher. In order to overcome the shortcomings of the premature convergence of the standard particle swarm optimization algorithm, an improved strategy for multiple particle swarm optimization algorithms is proposed. In order to improve the premature problem in the search performance algorithm of PSO algorithm, this paper combines the algorithm with the useful attributes of other algorithms to improve the particle diversity in the algorithm, enhance the global search ability of the particle, and achieve effective feature extraction. The research indicates that the method proposed in this paper has certain practical effects and can provide theoretical reference for subsequent related research.

Download Full-text