Volume 07,Issue 04

Ultrasonic System to Improve Position Measurement and Shape Recognition Based on Neural Network

Authors

Siamak Hadadi

Abstract
Today, with the advancement of technology, the use of remote detection systems in various fields of medicine, industry and robotics has become common. In this regard, accurate recognition of the position and shape of objects is one of the main concerns. Guidance systems for the blind, car parking, separation systems and intelligent robots are examples of these systems. The basis of these systems is based on one of the methods of image processing, infrared or ultrasound. In this paper, the researchers have dealt with a method that has high diagnostic accuracy at a low cost and is resistant to environmental conditions. The proposed method consists of two phases: position recognition and shape recognition. In the first phase, the exact position of the object is calculated using the laws of wave physics and the method known as the improved time of flight method. In the second phase, a set of ultrasonic sensors with a two-dimensional matrix arrangement and a neural network were used to detect the shape of the object. In the first phase, the average distance detection error decreased from 1.875, which was related to the time of flight method, to 0.271. In the second phase, the efficiency of the neural network was 96.8 and the error rate was 3.2. The system presented in both phases is very low cost and is associated with increased measurement accuracy, which can therefore be a good alternative to other methods.

Keyword: Ultrasound, Position recognition, Shape recognition, Time of flight, Neural network.

PDF [ 514.81 Kb ] | Endnote File | XML