27-08-2014, 12:49 PM
An Efficient Autoadaptive Edge Detection Approach for Flame and Fire Images Project
An Efficient Autoadaptive.pdf (Size: 162.88 KB / Downloads: 31)
Introduction
The characteristics of flame & fire are usually based on luminous nature. Characteristics & the
distributive aspect is visible via the frequency of flickering that depicts the geometrical measurements. These
measurements are important for various technologies to monitor the propagation of combustion flames with
effect from the present quality of fossil fuel, biomass & has been leading to a poor quality, instability of flames
due to combustion & high pollutant emissions in power generation plants[1].
The techniques for visualizing the characteristics of different flames of power generation industry &
laboratory research is emphasized using the prior electrical or optical sensing, ionization based detection &
thermocouple based detection to quantify the various parameters such as shape, size, stability etc [10]. Flames
are justified via the rapid variations over time being an important indication to the existence of flames. With the
volatility of flames, measured by its oscillating frequency > 0.5 of contours, chrominance & luminosity values
[8], it becomes an essential aspect to focus for the determination of the nature of fire/flame through its proper
edge detection.
A number of methods have been reported for the detection of the flame edges prior by trying to
represent in Fourier domain which does not carry any time information to be computed [8]. A flame edge
detection has been performed using Hidden Markov Models [7] using a multifunctional instrumentation system
for monitoring the characterization of combustion flames [2] & also by methods of subtracting the current image
from the background & detection by threshold [11]. Edge detection has also been achieved by using the
conventional methods of Sobel, Roberts, Lapliance, Prewitt, & Canny but the results are rather unclear with
discontinued edges & additional artifacts embedded in the outputs
Related Work
Edge detection is analyzed by using the mathematical representation of first order & second order
derivatives. The first order finds the gradient & second order gives the magnitude of the edge. A flame region
has a stronger luminance in comparison to its ambient background and the boundary between the flame region
and its background is mostly continuous. The strategy used is to detect the coarse and the superfluous edges in a
flame image, if there is only one main flame and if the image contains multiple flames, it is segmented as to
contain one flame. Identify the flame’s primary edge and remove irrelevant ones to project the continuous edge.
The common edge-detection methods like Sobel, Prewitt, Roberts, Canny and Laplacian method have
been applied with appropriate parameters to process typical flame images. Despite many parameters being finely
and appropriately adjusted in the use of these methods, flame edges could not be clearly identified. Figure 1(a)–
(f) shows examples of results obtained by the conventional edge-detection methods along with the original
image.
Proposed Work
As the algorithm discussed in the previous section of this paper, works for superfluous flames & unable
to detect the edges of fire/flame images with either low luminosity or low intensity, the proposed work corrects
the basic image by considering the individual components of any image - luminance, intensity & color.
The intensity factor is adjusted to lie between low & high thresholds for the entire image provided by a
color correction function. As the pixels that are below certain low levels of intensity are increased, this might
provide an adverse effect of glaring to objects contained in the image.
Hence a color correction is performed once again considering only the color component of the objects
changed to restore their prior levels
Results and Discussions
After implementing the algorithm as described in Section III, many of flame images were processed
using this algorithm so as to evaluate its effectiveness. Most of the flame images were taken for propane Bunsen
flames burning in open air. Some of the images were attained from the Internet with courtesy of permission of
use
Conclusion
After the flame characteristics are analyzed, a new flame edge-detection method has been developed
and evaluated in comparison with conventional methods. As per the existing algorithm for detection of fire &
flame edges giving a clear & continuous edges when compared to the other traditional methods that has been