## Flame front tracking by laser induced fluorescence spectroscopy and advanced image analysis

Optics Express, Vol. 8, Issue 5, pp. 278-287 (2001)

http://dx.doi.org/10.1364/OE.8.000278

Acrobat PDF (1418 KB)

### Abstract

This paper presents advanced image analysis methods for extracting information from high speed Planar Laser Induced Fluorescence (PLIF) data obtained from turbulent flames. The application of non-linear anisotropic diffusion filtering and of Active Contour Models (Snakes) is described to isolate flame boundaries. In a subsequent step, the detected flame boundaries are tracked in time using a frequency domain contour interpolation scheme. The implementations of the methods are described and possible applications of the techniques are discussed.

© Optical Society of America

## 1 Introduction

2. C.F. Kaminski, J. Hult, and M. Aldén, “High repetition rate planar laser induced fluorescence of OH in a turbulent non-premixed flame,” Appl. Phys. B **68**, 757–760 (2000). [CrossRef]

3. A. Dreizler, S. Lindenmaier, U. Maas, J. Hult, M. Aldén, and C.F. Kaminski, “Characterisation of a spark ignition system by planar laser-induced fluorescence of OH at high repetition rates and comparison with chemical kinetic calculations,” Appl. Phys. B **70**, 287–294 (2000). [CrossRef]

2. C.F. Kaminski, J. Hult, and M. Aldén, “High repetition rate planar laser induced fluorescence of OH in a turbulent non-premixed flame,” Appl. Phys. B **68**, 757–760 (2000). [CrossRef]

3. A. Dreizler, S. Lindenmaier, U. Maas, J. Hult, M. Aldén, and C.F. Kaminski, “Characterisation of a spark ignition system by planar laser-induced fluorescence of OH at high repetition rates and comparison with chemical kinetic calculations,” Appl. Phys. B **70**, 287–294 (2000). [CrossRef]

5. G.J. Smallwood, O.L. Gulder, D.R. Snelling, B.M. Deschamps, and I. Gokalp, “Characterization of flame front surfaces in turbulent premixed methane/air combustion,” Combustion and Flame **101(4)**, 461–470 (1995). [CrossRef]

6. R. Knikker, D. Veynante, J.C. Rolon, and C. Meneveau, “Planar Laser-Induced Fluorescence in a Turbulent Premixed Flame to analyze Large Eddy Simulation Models,” in *Proceedings of the 10th international Symposium on Turbulence, Heat and Mass Transfer*, Lisbon (2000). http://in3.dem.ist.utl.pt/downloads/lxlaser2000/pdf/26 3.pdf

7. B.D. Haslam and P.D. Ronney, “Fractal properties of propagating fronts in a strongly stirred fluid,” Phys. Fluids **7(8)**, 1931–1937 (1995). [CrossRef]

8. Y.-C. Chen and M.S. Mansour, “Topology of turbulent premixed flame fronts resolved by simultaneous planar imaging of LIPF of OH radical and rayleigh scattering,” Experiments in Fluids **26**, 277–287 (1999). [CrossRef]

9. O.L. Gulder, G.J. Smallwood, R. Wong, D.R. Snelling, R. Smith, B.M. Deschamps, and J.-C. Sautet, “Flame front surface characteristics in turbulent premixed propane/air combustion,” Combustion and Flame **120(4)**, 407–416 (2000). [CrossRef]

10. P. Perona and J. Malik, “Scale-space and edge detection using anisotropic diffusion,” IEEE Trans. on Pattern Analysis and Machine Intelligence **12(7)**, 629–639 (1990). [CrossRef]

11. M. Kass, A. Witkin, and D. Terzopoulos, “Snakes: Active Contour Models,” International Journal on Computer Vision **1(4)**, 321–331 (1988). [CrossRef]

## 2 Experimental setup

*Re*

_{T}<500). The principle of PLIF is to form a light sheet from a laser beam, using suitable optics, which traverses the flame. If the wavelength is tuned to match a molecular resonance line of OH then light from the sheet is inelastically scattered from the OH radicals present in the interaction region (see Fig. 1). This scattered light (fluorescence) is captured at right angles using a camera which is focused to image the illuminated flame cross-section. The local intensity in the recorded image is a function of the local OH concentration in the flame. Since OH is formed in the reaction zone of the flame and is rapidly quenched by cold unreacted gases, it is a good indicator of the flame front position in flames where the reaction zone is thin. In hot combusted gases, OH is removed more slowly, and a certain equilibrium concentration prevails depending on local temperatures and burnt gas composition. The purpose of the image processing techniques presented here is to extract the flame boundary marked by OH concentrations from such data as accurately as possible and to describe its dynamics as clearly as possible.

3. A. Dreizler, S. Lindenmaier, U. Maas, J. Hult, M. Aldén, and C.F. Kaminski, “Characterisation of a spark ignition system by planar laser-induced fluorescence of OH at high repetition rates and comparison with chemical kinetic calculations,” Appl. Phys. B **70**, 287–294 (2000). [CrossRef]

*A*

^{2}∑

^{+}←

*X*

^{2}Π electronic band. The laser output was passed through sheet forming optics before traversing the combustion system of interest. A combustion cell featuring two opposing tungsten electrodes was used to ignite mixtures of methane and air. Controlled degrees of turbulence could be imposed on the mixture via four high speed rotating fans.

## 3 Image processing stage

### 3.1 Preprocessing of raw image data

**70**, 287–294 (2000). [CrossRef]

### 3.2 Noise reduction using non-linear diffusion filtering

10. P. Perona and J. Malik, “Scale-space and edge detection using anisotropic diffusion,” IEEE Trans. on Pattern Analysis and Machine Intelligence **12(7)**, 629–639 (1990). [CrossRef]

## 4 Image analysis stage

### 4.1 Segmentation using Active Contour Models (ACM)

11. M. Kass, A. Witkin, and D. Terzopoulos, “Snakes: Active Contour Models,” International Journal on Computer Vision **1(4)**, 321–331 (1988). [CrossRef]

*s*)=(

*x*(

*s*),

*y*(

*s*)) where

*s*∊ [0, 1]. In a discrete setting the snake is defined as a set of

*N*nodes, v

_{i}(

*n*)=(

*x*

_{i}(

*n*),

*y*

_{i}(

*n*)) where

*x*

_{i}(

*n*) and

*y*

_{i}(

*n*) are the x- and y-coordinates of node

*n*at iteration

*i*,

*n*=1, 2, …,

*N*. Various forces act on the nodes yielding the following equation for updating their position

*ω*

_{1},

*ω*

_{2},

*ω*

_{3}and

*ω*

_{4}are scalar factors weighting the different forces that we incorporate to deform the snake during its iterations [15

15. T. McInerney and D. Terzopoulos, “T-Snakes: Topology adaptive snakes,” Medical Image Analysis **4**, 73–91 (2000). [CrossRef] [PubMed]

*n*) is a tensile force, which resists stretching of the snake, acting on node

*n*at iteration

*i*and is given in discrete form as

*n*) is a flexural force which resists bending of the snake and is given as

*n*) is an inflation force designed to move the snake nodes in a direction normal to the contour they form. In the cases where the snake is a closed contour, as in our application images, this means the nodes will move inwards or outwards. This will either inflate or deflate the snake towards the target boundary which enables us to initialize the snake at locations far from the target object that we want to segment.

*n*) is defined as

**n**

_{i}(

*n*) is the unit vector in the direction normal to the contour at node

*n*and

*I*

_{s}(

*x*

_{i}(

*n*),

*y*

_{i}(

*n*)) is the intensity of the pixel (

*x*

_{i}(

*n*),

*y*

_{i}(

*n*)) in a smoothed version of the image. The binary function

*T*is an image intensity threshold.

*n*) is an external force that is derived from the image data in a way that causes the snake nodes to move towards regions of higher intensity gradient (mainly edges) in the image and is defined as

*P*is the image gradient reflecting high intensity changes commonly present at boundary points.

16. S. Lobregt and M. Viergever, “A discrete dynamic contour model,” IEEE Trans. on Medical Imaging **14(1)**, 12–24 (1995). [CrossRef]

### 4.2 Temporal interpolation of the flame contours

*j*we have a snake contour {v (

*n*,

*j*)=(

*x*(

*n*,

*j*),

*y*(

*n*,

*j*)),

*n*=1, 2, …,

*N*} for

*j*=1, 2, …,

*F*where

*F*is the number of frames (which in the present case is 4). In order to interpolate, we start by re-parameterizing each of the original flame contours with a new shape representation. This is most efficiently done by transforming from the spatial into the frequency domain. An advantage is that the need for a node-to-node correspondence between different contours (snakes) is avoided. The one-dimensional discrete cosine transform (DCT) of the sequence of

*x*(

*n*,

*j*) contour coordinates (and similarly for the

*y*(

*n*,

*j*) coordinates),

*n*=1, 2, …,

*N*, is defined as

*k*=1, 2, …,

*N*. Using the DCT as a new frequency domain shape parameterization has many advantages: It produces real coefficients, has excellent energy compaction properties, as well as having coefficients which correspond (opposed to spatial contour points with no point-to-point correspondence). Now armed with these frequency coefficients as new curve parameters, we can directly perform the actual interpolation. In our implementation cubic spline interpolation between corresponding frequency coefficients was utilized. Finally the Inverse Discrete Cosine Transform (IDCT) is used to transform the interpolated components back into the spatial domain:

*n*=1, 2, …,

*N*and

*j′*spans the interpolated frames (including the original ones).

*F*shapes and each shape contains

*L*nodes. Both the

*x*and

*y*coordinates of each node move throughout the sequence (in time) according to sinusoidal functions with different amplitudes and frequencies, which causes spatial shape deformations in time. The coordinates are also scaled differently between frames according to sinusoidal functions in order to produce dynamic shapes that shrink and expand with time. To quantify the error (difference) between the original (known) synthetic sequences and the interpolated ones, we define the following error measure for each shape in the sequence

*A*

_{o}and

*A*

_{i}are the areas enclosed within the original and interpolated shapes, respectively. Fig. 5 illustrates some of these test examples and reports the corresponding average error values (over all frames in each test sequence) between the original and the interpolated sequences. Fig. 6 shows some further qualitative testing examples.

### 4.3 Flame front velocity estimation

## 5 Discussion and Conclusions

18. J. Hult, G. Josefsson, M. Aldén, and C.F. Kaminski, “Flame front tracking and simultaneous flow field visualization in turbulent combustion,” in *Proceedings of the 10th International Symposium on Application of Laser Techniques to Fluid mechanics*, Lisbon (2000). http://in3.dem.ist.utl.pt/downloads/lxlaser2000/pdf/26 2.pdf

15. T. McInerney and D. Terzopoulos, “T-Snakes: Topology adaptive snakes,” Medical Image Analysis **4**, 73–91 (2000). [CrossRef] [PubMed]

## References and links

1. | J. Warnatz, U. Maas, and R.W. Dibble, |

2. | C.F. Kaminski, J. Hult, and M. Aldén, “High repetition rate planar laser induced fluorescence of OH in a turbulent non-premixed flame,” Appl. Phys. B |

3. | A. Dreizler, S. Lindenmaier, U. Maas, J. Hult, M. Aldén, and C.F. Kaminski, “Characterisation of a spark ignition system by planar laser-induced fluorescence of OH at high repetition rates and comparison with chemical kinetic calculations,” Appl. Phys. B |

4. | J. Hult, A. Omrane, J. Nygren, C.F. Kaminski, B. Axelsson, R. Collin, P.-E. Bengtsson, and M. Aldén, “Quantitative three dimensional imaging of soot volume fraction in turbulent non-premixed flames”, (in preparation). |

5. | G.J. Smallwood, O.L. Gulder, D.R. Snelling, B.M. Deschamps, and I. Gokalp, “Characterization of flame front surfaces in turbulent premixed methane/air combustion,” Combustion and Flame |

6. | R. Knikker, D. Veynante, J.C. Rolon, and C. Meneveau, “Planar Laser-Induced Fluorescence in a Turbulent Premixed Flame to analyze Large Eddy Simulation Models,” in |

7. | B.D. Haslam and P.D. Ronney, “Fractal properties of propagating fronts in a strongly stirred fluid,” Phys. Fluids |

8. | Y.-C. Chen and M.S. Mansour, “Topology of turbulent premixed flame fronts resolved by simultaneous planar imaging of LIPF of OH radical and rayleigh scattering,” Experiments in Fluids |

9. | O.L. Gulder, G.J. Smallwood, R. Wong, D.R. Snelling, R. Smith, B.M. Deschamps, and J.-C. Sautet, “Flame front surface characteristics in turbulent premixed propane/air combustion,” Combustion and Flame |

10. | P. Perona and J. Malik, “Scale-space and edge detection using anisotropic diffusion,” IEEE Trans. on Pattern Analysis and Machine Intelligence |

11. | M. Kass, A. Witkin, and D. Terzopoulos, “Snakes: Active Contour Models,” International Journal on Computer Vision |

12. | C.F. Kaminski, J. Hult, M. Aldén, S. Lindenmaier, A. Dreizler, U. Maas, and M. Baum, “Complex turbulence/chemistry interactions revealed by time resolved fluorescence and direct numerical simulations,” Proc. Combust. Inst. 28, The Combustion Institute, Pittsburgh, in press (2000). |

13. | F. Catté, P.-L. Lions, J.-M. Morel, and T. Coll, “Image selective smoothing and edge detection by nonlinear diffusion,” SIAM J. Numer. Anal. |

14. | H. Malm, J. Hult, G. Sparr, and C.F. Kaminski, “Non-linear diffusion filtering of images obtained by planar laser induced florescence spectroscopy,” JOSA A |

15. | T. McInerney and D. Terzopoulos, “T-Snakes: Topology adaptive snakes,” Medical Image Analysis |

16. | S. Lobregt and M. Viergever, “A discrete dynamic contour model,” IEEE Trans. on Medical Imaging |

17. | A. Jain, |

18. | J. Hult, G. Josefsson, M. Aldén, and C.F. Kaminski, “Flame front tracking and simultaneous flow field visualization in turbulent combustion,” in |

19. | V. Caselles, R. Kimmel, and G. Sapiro, “Geodesic active contours,” in |

**OCIS Codes**

(100.0100) Image processing : Image processing

(100.2960) Image processing : Image analysis

(100.2980) Image processing : Image enhancement

(300.6280) Spectroscopy : Spectroscopy, fluorescence and luminescence

**ToC Category:**

Research Papers

**History**

Original Manuscript: December 19, 2000

Published: February 26, 2001

**Citation**

Rafeef Abu-Gharbieh, Ghassan Hamarneh, Thomas Gustavsson, and Clemens Kaminski, "Flame front tracking by laser induced fluorescence spectroscopy and advanced image analysis," Opt. Express **8**, 278-287 (2001)

http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-8-5-278

Sort: Journal | Reset

### References

- J. Warnatz, U. Maas, and R.W. Dibble, Combustion - physical and chemical fundamentals, modeling and simulation, experiments, pollutant formation (Springer-Verlag, Heidelberg 1996).
- C.F. Kaminski, J. Hult, and M. Ald�n, ``High repetition rate planar laser induced fluorescence of OH in a turbulent non-premixed flame,' Appl. Phys. B 68, 757-760 (2000). [CrossRef]
- A. Dreizler, S. Lindenmaier, U. Maas, J. Hult, M. Ald�n, and C.F. Kaminski, ``Characterisation of a spark ignition system by planar laser-induced fluorescence of OH at high repetition rates and comparison with chemical kinetic calculations,' Appl. Phys. B 70, 287-294 (2000). [CrossRef]
- J. Hult, A. Omrane, J. Nygren, C.F. Kaminski, B. Axelsson, R. Collin, P.-E. Bengtsson, and M. Ald�n, ``Quantitative three dimensional imaging of soot volume fraction in turbulent non-premixed flames', (in preparation).
- G.J. Smallwood, O.L. Gulder, D.R. Snelling, B.M. Deschamps, and I. Gokalp, ``Characterization of flame front surfaces in turbulent premixed methane/air combustion,' Combustion and Flame 101(4), 461-470 (1995). [CrossRef]
- R. Knikker, D. Veynante, J.C. Rolon, and C. Meneveau, ``Planar Laser-Induced Fluorescence in a Turbulent Premixed Flame to analyze Large Eddy Simulation Models,' in Proceedings of the 10th international Symposium on Turbulence, Heat and Mass Transfer, Lisbon (2000), http://in3.dem.ist.utl.pt/downloads/lxlaser2000/pdf/26\_3.pdf
- B.D. Haslam and P.D. Ronney, ``Fractal properties of propagating fronts in a strongly stirred fluid,' Phys. Fluids 7(8), 1931-1937 (1995). [CrossRef]
- Y.-C. Chen and M.S. Mansour, ``Topology of turbulent premixed flame fronts resolved by simultaneous planar imaging of LIPF of OH radical and rayleigh scattering,' Experiments in Fluids 26, 277-287 (1999). [CrossRef]
- O.L. Gulder, G.J. Smallwood, R. Wong, D.R. Snelling, R. Smith, B.M. Deschamps, and J.-C. Sautet, `` Flame front surface characteristics in turbulent premixed propane/air combustion,' Combustion and Flame 120(4), 407-416 (2000). [CrossRef]
- P. Perona and J. Malik, ``Scale-space and edge detection using anisotropic diffusion,' IEEE Trans. on Pattern Analysis and Machine Intelligence 12(7), 629-639 (1990). [CrossRef]
- M. Kass, A. Witkin, and D. Terzopoulos, ``Snakes: Active Contour Models,' International Journal on Computer Vision 1(4), 321-331 (1988). [CrossRef]
- C.F. Kaminski, J. Hult, M. Ald�n, S. Lindenmaier, A. Dreizler, U. Maas, and M. Baum, ``Complex turbulence/chemistry interactions revealed by time resolved fluorescence and direct numerical simulations,' Proc. Combust. Inst. 28, The Combustion Institute, Pittsburgh, in press (2000).
- F. Catt�, P.-L. Lions, J.-M. Morel, and T. Coll, ``Image selective smoothing and edge detection by nonlinear diffusion,' SIAM J. Numer. Anal. 29, 182-193 (1992). [CrossRef]
- H. Malm, J. Hult, G. Sparr, and C.F. Kaminski, ``Non-linear diffusion filtering of images obtained by planar laser induced florescence spectroscopy,' J. Opt. Soc. Am. A 17, 2148-2156 (2000). [CrossRef]
- T. McInerney and D. Terzopoulos, ``T-Snakes: Topology adaptive snakes,' Medical Image Analysis 4, 73-91 (2000). [CrossRef] [PubMed]
- S. Lobregt and M. Viergever, ``A discrete dynamic contour model,' IEEE Trans. on Medical Imaging 14(1), 12-24 (1995). [CrossRef]
- A. Jain, Fundamentals of digital image processing (Prentice Hall, 1989).
- J. Hult , G. Josefsson, M. Ald\'{en, and C.F. Kaminski, ``Flame front tracking and simultaneous flow field visualization in turbulent combustion,' in Proceedings of the 10th International Symposium on Application of Laser Techniques to Fluid mechanics, Lisbon (2000), http://in3.dem.ist.utl.pt/downloads/lxlaser2000/pdf/26\_2.pdf
- V. Caselles, R. Kimmel, and G. Sapiro, ``Geodesic active contours,' in Proceedings of the International Conference on Computer Vision, 694 -699 (1995).

## Cited By |
Alert me when this paper is cited |

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.

OSA is a member of CrossRef.