Skip to Content

Image Recognition

A Supervised Approach to Electric Tower Detection and Classification for Power Line Inspection

Publication
A Supervised Approach to Electric Tower Detection and Classification for Power Line Inspection
Inspection of power line infrastructures must be periodically conducted by electric companies in order to ensure reliable electric power distribution. Research efforts are focused on automating the power line inspection process by looking for strategies that satisfy the different requirements of the inspection: simultaneously detect transmission towers, check for defects, and analyze security distances. Following this direction, this paper proposes a supervised learning approach for solving the tower detection and classification problem, where HOG features are used to train two MLP (multi-layer perceptron) neural networks. The first classifier is used for backgroundforeground separation, and the second multi-class MLP is used for classifying 4 different types of electric towers. A thorough evaluation of the tower detection and classification approach has been carried out on image data from real inspections tasks with different types of towers and backgrounds. In the different evaluations that were conducted highly encouraging results were obtained. This shows that a learning-based approach is a promising technique for power line inspection.
Congresses name: 

International Joint Conference on Neural Networks (IJCNN 2014)

Place: 

Beijing

Date: 
July 6-11

Abstract:

Visual Identification and Tracking for Vertical and Horizontal Targets in Unknown Indoor Environment

Publication
Visual Identification and Tracking for Vertical and Horizontal Targets in Unknown Indoor Environment
This paper presents the visual identification and tracking module developed for automatic character recognition on-board Unmanned Aerial Vehicles(UAVs) to be used in the IMAV 2012 "Indoors Autonomy Competition". The presented strategy is based on three stages: image segmentation, visual tracking and optical character recognition(OCR). The module is validated in a house-like environment, which has the same characteristic as the ones used in the competition, including challenging distracters such as windows, door and illumination that represent a challenging scenario for the automatic character recognition sub-task. The results show that each stage in this module can accomplish its function successfully and obtain the correct information from the vertical and horizontal targets, based on the obtained visual information and on-board sensors.
Congresses name: 

International Micro Air Vehicle Conference and Flight Competition, IMAV 2012

Place: 

Braunschweig, Germany

Date: 
3-6 July 2012

This paper presents the visual identification and tracking module developed for automatic character recognition on-board Unmanned Aerial Vehicles(UAVs) to be used in the IMAV 2012 "Indoors
Autonomy Competition". The presented strategy is based on three stages: image segmentation, visual tracking and optical character recognition(OCR). The module is validated in a house-like environment, which has the same characteristic as the ones used in the competition, including challenging distracters such as windows, door and illumination that represent a challenging scenario for the automatic character recognition sub-task. The results show that each stage in this module can accomplish its function successfully and obtain the correct information from the vertical and horizontal targets, based on the obtained visual information and on-board sensors.



Fig 1. Manual test flight with the Asctec Pelican quadrotor on the house replica of the IMAV 2012 competition.


Character recognition example
Fig 2. The total process of image segmentation. The left process shows the white ROI and its detection result, and the right process shows the black ROI and its recognition result.

Our team could not finally take part in the "Indoors Autonomy" challange. However, the Computer Vision Group team participated in the "Indoor Flight Dynamics – Rotory Wing MAV" challenge and obtained two prizes: the "Best Automatic Performance - IMAV 2012" award and the second position in this challenge (related link) .

"Real-time recognition of patient intentions from sequences of pressure maps using artificial neural networks"

Journal: 
Computers in Biology and Medicine, ELSEVIER

ISSN: 0010-4825

Paper reference: 
Vol 42, Issue 4 , Pages: 364–375, DOI http://dx.doi.org/10.1016/j.compbiomed.2011.12.003
Publication Date: 
APRIL 2012

 Abstract

Objective: In this paper we address the problem of recognising the movement intentions of patients restricted to a medical bed. The developed recognition system will be used to implement a natural human–machine interface to move a medical bed by means of the slight movements of patients with reduced mobility.

A Visual AGV-urban Car using Fuzzy Control

An autonomous visual line guided urban-car using no prediction information and controlled with Fuzzy Logic
This is an autonomous urban-car, Citroën C3, guided with the visual information obtained by a camera installed in front of the car.A Fuzzy controller was design to control the steering wheel of the vehicle. The only input information for the controller is the visual distance from the centre of the image and the centre of the line. No more information from other sensor and previous information about the path are used.
Views

 

Dimensionality Reduction by Self Organizing Maps that preserve distances in the Output Space

Congresses name: 

International Joint Conference on Neural Networks IJCNN09

Place: 

Atlanta - U.S.A.

Date: 
14-19 June 2009
Authors: 

Abstract:

Syndicate content