Masters Theses

Date of Award

8-1991

Degree Type

Thesis

Degree Name

Master of Science

Major

Electrical Engineering

Major Professor

L. Montgomery Smith

Committee Members

Roger Crawford, Bruce Bomar

Abstract

A system for detecting gas leaks is described and tested. The algorithm used by the system detects the occurrence of an abrupt but sustained change in a time sequence of video images. The system consists of a high pass filter cascaded with a moving average filter. The highpass filter removes slowly varying background intensity values and its output is averaged with a set number of previous outputs by the moving average. The absolute value of the average is taken and then compared to a threshold to decide if a step-like change has occurred. It is shown that for a chosen cutoff frequency in the high pass filter, an optimal value for the number of terms in the moving average exists. The cutoff frequency and the corresponding optimal number of terms in the average, as well as, the threshold value determine the smallest step amplitude that can be detected.

The algorithm was tested in a numerical study by implementing it in a FOR­ TRAN program on computer generated input data. The results are shown as the probability of correct detection versus step size and illustrate the effects of noise on the input and the value of the threshold on the performance of the algorithm. Next, the algorithm was tested on a sequence of images of an actual gas leak. A gas leak scene was set up in the laboratory and sequential image data was acquired. The data was processed on an image processor where the algorithm was implemented at every point in an image. The results are shown as binary output images in which white regions indicate areas where change was detected. The parameters of the algorithm were varied and the results examined. The results presented show the effects of the parameters on the size of the detected region as well as the noise sensitivity of the detector. These results verify that the parameters can be chosen such that this is an effective method for leak detection.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS