Computer Vision Fire Safety Automation Neuro-Fuzzy Algorithms

Machine Vision Based Fire Flame Detection Using Multi-Features and Autonomous Fire Extinguishing System Vision-Based Fire Detection in Videos
Dr.KeiraGibbs Profile Pic
Dr.KeiraGibbs,United States,Researcher
Published Date:12-11-2017
Your Website URL(Optional)
I.J. Information Technology and Computer Science, 2015, 04, 14-27 Published Online March 2015 in MECS ( DOI: 10.5815/ijitcs.2015.04.02 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms Manjunatha K.C. M/s. Prakash Sponge Iron and Power (P) Ltd (ERM Group), Chitradurga - 577501, India Email:, Dr. Mohana H.S, Dr. P.A Vijaya Professor, Dept. of IT & EC Malnad College of Engineering, Hassan - 573201, India Email:, Abstract- A computer vision-based automated fire detection may become out of reach of the sensors and can reduce and suppression system for manufacturing industries is the possibility of detection. presented in this paper. Automated fire suppression system Vision-based fire detection and automated suppression plays a very significant role in Onsite Emergency System (OES) system offers several advantages. First, the cost is less, as as it can prevent accidents and losses to the industry. A rule this system is based on cameras and industries are mostly based generic collective model for fire pixel classification is equipped with CCTVs for surveillance. SCADA and proposed for a single camera with multiple fire suppression PLC may also be present if there is process automation. chemical control valves. Neuro-Fuzzy algorithm is used to Second, the response time is faster as it does not have to identify the exact location of fire pixels in the image frame. Again the fuzzy logic is proposed to identify the valve to be wait for any product of combustion. Fire suppression controlled based on the area of the fire and intensity values of chemical control valve will be operational at the start of the fire pixels. The fuzzy output is given to supervisory control fire itself thus reducing the scope for spreading of fire. and data acquisition (SCADA) system to generate suitable Finally, in case of false alarm, confirmation can be done analog values for the control valve operation based on fire from a control room without rushing to the location. As it characteristics. Results with both fire identification and is important to have a fast fire detection and suppression suppression systems have been presented. The proposed method system, a computer vision based technique is proposed in achieves up to 99% of accuracy in fire detection and automated this paper. This paper initially focuses on video and suppression. image processing for flame pixels detection. Once the Index Terms- Onsite Emergency System, SCADA, PLC, fire is confirmed the focus is on the computation of Weighted Centroid, Fire Pixel Number, Neuro-Fuzzy location and intensity of the fire using neuro-fuzzy Algorithm. algorithms. Finally, the attention is on the fire suppression chemical control valve operation through SCADA, PLC and I/O configuration. In industries, I. INTRODUCTION critical areas which are prone to fire accidents are identified by a survey. CCD cameras are installed in Vision-based fire detection and automated suppression these areas with proper scene planning. Also a number of (VFDAS) systems are one of the most important fire suppression chemical control valves are installed mechanisms in manufacturing and process industries. It which are connected to centralized chemical pumping is more critical in industries which use oil, gas and station. Series of image frames from the continuous video petrochemicals as fuels. A fast automated detection stream is acquired. These frames are processed with a system has to be ready in order to prevent any fire stored image taken when the situation is normal - accidents and avoid loss of life and property. VFDAS is considered as the background image. Successive frames newly developed technique based on computer vision, th th (i and i+1 ) of the video stream and normal background image processing and neuro-fuzzy algorithms. Vision- image (j) are processed at a time. Fire rules are applied based fire detection (VFD) has many advantages over to identify the fire pixels. In second stage, the particular traditional methods such has fast response, non-contact location where fire has occurred is identified by using and no installation limitations. Currently, most of the fire neural-network. Finally, the fire suppression chemical detection systems use sensors for detecting smoke, rise in control valve in that location is operated with the help of temperature etc. They need a considerable time for fuzzy logic. The control system communication is done responding as the sensors require product of fire (e.g., through PROFIBUS and I/O signal conditioning circuits. smoke, temperature etc.) to reach the sensors. Also they This paper is structured in such a way that literature have to be carefully placed in selected locations. Such a pertaining to recent developments in computer vision and sensor-based fire detection system is unsuitable for large their applications. Problem is defined based on industries with open spaces when products of combustion Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 15 Automation by Using Neuro-Fuzzy Algorithms fundamental study and practical experience, the fire carried out for smoldering fire SH1 and flaming fire SH3 characteristics are studied in profound manner for the of the china national standard test fires. The simulation implementation. Methodologies are framed for the study shows that the model combines the advantages of automated fire suppression system and sequential fuzzy system and neural network, and improves the implementation has been done to achieve the anticipated intelligence of fire detection, has a stronger ability to results. All the sequential results are presented and adapt the environment. GUO Jian, ZHU Jie, and ZHAO discussed in the vision of performance of the Mingru (2009) are applied Self-Adaptive Neural Fuzzy implemented system. Network in Early Detection of Conveyor Belt Fire. It has four inputs to neural fuzzy called temperature, rate of temperature change, dense of carbon monoxide and rate II. RELATED WORKS of CO dense change. In 2004, the fire detection method proposed by Chen et al. adopted the RGB color based Several approaches have been suggested in literature to chromatic model and used disorder measurement. They identify fire by using variety of image processing used the intensity and saturation of red component and techniques. But very few systems have been attempted the segmentation by image differencing. In 2007, Lee and for automated fire suppression. Tao Chen, Hongyong Han also used the RGB color input video for real-time Yuan, Guofeng Su, Weicheng Fan (2004) experimented fire detection in the tunnel environment with many an automatic fire searching and suppression system with predetermined threshold values. remote controlled fire monitors for large space. The fire searching method is based on computer vision using one CCD camera fixed at the end of a fire monitor chamber. III. PROBLEM DEFINITION A sensor based automated firefighting system with smoke and temperature detection has been attempted by Fire is a big threat for large and medium scale Mohammad Jane Alam Khan Muhammed Rifat Imam, industries involved in production or processing. Fire Jashim Uddin, M. A. Rashid Sarkar (2012). An accidents result in loss of life and property. It damages autonomous fire extinguishing system that detects, goodwill and effects environment severely. It may affect targets, and extinguishes a fire within a working space by the industries around it. A survey conducted by using heat sensors, a flow control system, servo motors, Federation of India Chambers Commerce and Industry and a water extinguishing gun has been implemented by (FICCI) in 2010-11, has reported as many as 22,187 fire A. Rehman, N. Masood, S. Arif, U. Shahbaz (2012). related calls resulting in 447 deaths and 2,613 injuries Andrey N. Pavlov, Evgeniy S. Povemov (2009), across India. The risk of fire has been rated among the conducted an experiment to test of automatic fire gas top six risks in India. In a survey conducted by Allianz explosion suppression system. Changwoo Ha, Ung Global Corporate & Specialty (AGCS) during January Hwang, Gwanggil Jeon, Joongwhee Cho, and Jechang 2013, it was found that fire and explosion replaced Jeong (2012) proposed vision-based fire detection “economic risk” as the third most important forward- algorithm by using optical flow algorithm. Tian Qiu, looking risk for the year ahead. Fig (1) shows fire Yong Yan and Gang Lu (2012) have attempted the explosions in different industrial areas along with the determination of flame or fire edges. A rule-based current suppression system. Fire occurred at motor generic color model for flame pixel classification has control center of a steel plant due to short circuit shown been proposed by Turgay C- elik, Hasan Demirel (2008). in figs 1(a) and 1(b). Fig 1(c) shows the fire caused due Bo-Ho Cho, Jong-Wook Bae, and Sung-Hwan Jung to improper closing of manhole door in the steel making (2008) studied about automatic fire detection system furnace. Fire on electrical cable tray due to overheat is without the heuristic fixed threshold values and presented shown in fig 1(d). A fire explosion at power plant an automatic method using the statistical color model and transformer station is shown in figs 1(e) and 1(f). Fig 1(g) the binary background mask. A Multisensor Fire and 1(h) show a fire explosion in a natural gas and oil detection algorithm is suggested by KuoL. Su (2006). refinery industry. Currently available manual fire Hideaki Yamagishi Jun‟ichi Yamaguchi (1999) suppression systems are shown in figs 1(i) and 1(j). developed a method which fire flame can be detected by calculating a space-time fluctuation data on a contour of the flame area extracted by color information. A hybrid clustering algorithm for fire detection is proposed by Ishita Chakraborty and Tanoy Kr. Paul (2010), the same analyzed with color based thresholding method. Dengyi Zhang, Shizhong Han, Jianhui Zhao (2009) proposed a real-time forest fire detection algorithm using artificial neural networks based on dynamic characteristics of fire regions segmented from video images. Fire region is obtained from image with the help of threshold values in HSV color space. A fire detection model based on fuzzy neural network is studied by Quanmin GUO Junjie DAI Jian WANG (2010) and the simulation experiments were Fig.1.(a) Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 16 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms Fig.1.(b) Fig.1.(f) Fig.1.(g) Fig.1.(c) Fig.1.(h) Fig.1.(d) Fig.1.(e) Fig.1.(i) Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 17 Automation by Using Neuro-Fuzzy Algorithms Fig. 3.(c) Fig.1.(j) Fig. 1. (a) & (b) Fire at electrical control panels of steel plant, (c) Improper closing of steel furnace door caused fire leakage, (d) Fire in electrical cable tray, (e) & (f) Fire at power plant substation, (g) Fire explotion at natural gas plant, (h) Fire explosion at oil refinary industry, (i) & (j) Currently available manual fire suppression system. Fig. 3.(d) Fig. 2. Avearge RGB values of fire region. Fig. 3.(e) Fig. 3.(a) Fig. 3.(f) Fig. 3. (b) RGB intensity values of circle marked pixels in fire like sodium bulb image (a). (d) RGB intensity values of circle marked pixels in fire image (c). (f) RGB intensity values of circle marked pixels in fire Fig. 3.(b) image. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 18 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms IV. CHARECTERSTICS OF FIRE (5) It is well known that fire has unique visual signatures. B. Image Subtraction Colour, geometry, and motion of fire region are features The subtraction of two images is performed in a single for efficient classification. The shape of fire region often pass. The output pixel values are given by: keeps changing and exhibits a stochastic motion, which depends on surrounding environmental factors such as Q(i, j) = P (i, j) – P (i, j) (6) 1 2 the type of burning elements and wind. Major useful Where P & P are input images and Q is the output features for detecting fire are colour, randomness of fire 1 2 image with i columns and j rows. area size fire boundary roughness, surface coarseness, skewness and spatial distribution. Fire has very distinct C. Binary Conversion. colour characteristics, and although empirical, it is the Binary images are obtained from gray-scale images by most powerful single feature for finding fire in video thresholding operations. A thresholding operation sequences. chooses some of the pixels in the image as foreground Based on tests with several images in different pixels that make up the objects of interest and rest as resolutions and scenarios, it is reasonably assumed that background pixels. Binary images are simplest to generally the color of flames belongs to the red-yellow process and outline of the object is easily obtainable. range. Laboratory experiments show that this is the case  Binary transformation map b : C → 0,1 for hydrocarbon flames by C. E. Baukal, Jr. (2001) which  Thresholded image b ◦ I : X → 0, 1 are the most common type of flames seen in nature. For  For each x ∈ X, these type of flames it is noticed that for a given fire pixel, the value of red channel is greater than the green channel, (7) and the value of the green channel is greater than the value of blue channel, as illustrated in Fig 2. D. Impulse noise and Median filtering RGB intensity values of specifically earmarked pixels of the image have been plotted in the fig 3. Fire-like The PDF of impulse noise is given by images and their pixels intensity values are plotted in figs 3(a) and 3(b). Fire images and their various regions pixel (8) values are plotted in figs 3(c), 3(d), 3(e) and 3(f). Noise impulses can be negative or positive. Negative impulses appear as black (pepper) points and positive V. METHODOLOGIES impulses appear white (salt) noise. A. RGB to HSV Conversion Median Filtering: The best known filter for noise In this work, RGB data obtained from a color camera reduction is the median filter, which replaces the value of is transformed into HSV data. In the HSV color space, a pixel by the median of the gray levels in the gray scale and color information are in separate channels. neighborhood of that pixel- Then, if a pixel‟s color is transformed into the flame ̂ (9) color region in the HSV space, the pixel is regarded as a flame color area in the input image. The color The original value of the pixel is included in the information is used for quick attention. This method does computation of the median. not suffer from the contrast distortion issues seen in RGB-based combination. E. Decision Rules for fire pixels RGB to HSV color space conversion can be done First phase of the proposed fire-detection algorithm is using the following set of equations: based on RGB color model. The hue value of fire pixels is in the range of 0 to 60 and the corresponding RGB V = Max (R,G,B) (1) value will be mapped to the conditions: RG and GB, i.e., the color range of red to yellow. In spite of various (2) colors of fire flames, the initial flame frequently displays red-to-yellow color. Thus, the condition to detect fire If S = 0, Then H = 0; color in the proposed method is RGB. The value of R If R = V, Then component should be over a threshold R . To avoid the T effect of background illumination, the saturation value of fire-flame is set to be over some threshold. (3) Three rules are used for identifying fire pixels in an image: If G = V, Then Rule 1: R R T (4) Rule 2: R ≥ G B Rule 3: IF (S ≥ ((255 – R) × S / R )) (10) T T If B = V, Then Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 19 Automation by Using Neuro-Fuzzy Algorithms FIRE = ((Rule1 AND Rule 2) OR Rule 3) (11) ∑ (13) ∑ (14) ∑ (15) where (x , y ) is the spatial location of the pixel, Y , i i mean Cb , and Cr are the mean values of luminance, mean mean Chrominance Blue, and chrominance Red channels of pixels, and K is the total number of pixels in image. The rules defined for RGB colour space, i.e. R≥G≥B, and R≥R can be translated into YCbCr space as mean Y(x,y) ˃ Cb(x,y) (16) Cr (x,y) ˃ Cb (x,y) (17) where Y(x,y), Cb(x,y), and Cr(x,y) are luminance, Fig. 4. Relation between R component and saturation. Chrominance-Blue and Chrominance Red values at the spatial location (x, y). 16 and 17 imply, respectively, that In rule 3, S denotes the value of saturation when the T flame luminance should be greater than Chrominance value of R component is R for the same pixel. Based on T Blue and Chrominance Red should be greater than the the concept that the saturation will degrade with the increasing R component, we have the expression ((255- Chrominance Blue. 16 and 17 can be interpreted to be a consequence of the fact that the flame has saturation in R)S /R ). The relation between R component and T T Saturation for the extracted fire pixels is shown in the Fig red colour channel (R). For the fire pixels Y colour value is greater than Cb colour value and Cr colour value is 4. greater than the Cb colour value Turgay C-elik, Hasan In the second phase fire pixel classification is done for Demirel (2008). which YCbCr colour space is better. RGB colour space has disadvantages of illumination dependence. Besides these two rules, since the flame region is generally the brightest region in the observed scene, the Furthermore, it is not possible to separate a pixel‟s value into intensity and chrominance. The chrominance can be mean values of the three channels, in the overall image - Y , Cb , and Cr - contain valuable information. used in modelling colour of fire rather than its intensity. mean mean mean The conversion from RGB to YCbCr colour space is For the flame region the value of the Y component is larger than the mean Y component of the overall image formulated as follows. while the value of Cb component is in general smaller than the mean Cb value of the overall image. = + (12) Furthermore, the Cr component of the flame region is larger than the mean Cr component. These observations Where Y is Luminance, Cb and Cr are Chrominance have been verified in a large number of experiments with Blue and chrominance Red components, respectively. images containing fire regions and are formulated as the The range of Y is 16 235, Cb and Cr are equal to 16 following rule: 240. For a given image, one can define the mean values of the three components in YCbCr colour space as (18) Any pixel F(x, y) which satisfies condition in Equation  Compute gradient of g (m,n) using any of the (18) is labelled as fire pixel. gradient operators (Roberts, Sobel, Prewitt, etc) to get: F. Edge Detection Edge detection is done using the procedure by Canny. √ (21) The purpose is to detect edges suppressing the noise at and the same time.  Smoothen the image with a Gaussian filter to reduce (22) noise and unwanted details and textures.  Threshold M: g(m,n) = G (m,n) f(m,n) (19) σ (23) where where T is so chosen that all edge elements are kept (20) √ while most of the noise is suppressed Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 20 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms  Suppress non-maxima pixels in the edges in M T obtained above to thin the edge ridges (as the edges might have been broadened in step a. To do so, check to see whether each non-zero M (m,n) is T greater than its two neighbors along the gradient direction θ (m, n). If so, keep M (m,n) unchanged, T otherwise, set it to 0.  Threshold the previous result by two different values t and t (where t t ) to obtain two binary 1 2 1 2 Fig. 5. Neural network system. images T and T . Compared to T , T has less noise 1 2 1 2 and fewer false edges but larger gaps between edge Fuzzy logic has two different meanings. In a narrow segments. sense, fuzzy logic is a logical system, which is an  Link the edge segments in T to form continuous 2 extension of multi-valued logic. However, in a wider edge. To do so, trace each segment in T to its end 2 sense fuzzy logic (FL) is almost synonymous with the and then search its neighbors in T to find any edge 1 theory of fuzzy sets, a theory which relates to classes of segment in T to bridge the gap until reaching 1 objects with unsharp boundaries in which membership is another edge segment in T . 2 a matter of degree. In this perspective, fuzzy logic in its G. Boundary Detection and Fire Area Calculations. narrow sense is a branch of FL. Even in its more narrow The Moore-Neighbor tracing algorithm modified by definition, fuzzy logic differs both in concept and Jacob‟s stopping criteria has been explored for boundary substance from traditional multi-valued logical systems. detection. Using Jacob's stopping criterion will greatly The point of fuzzy logic is to map an input space to an improve the performance of Moore-Neighbor tracing output space, and the primary mechanism for doing this making it the best algorithm for extracting the contour of is a list of if-then statements called rules. All rules are any pattern irrespective of its connectivity. The reason evaluated in parallel, and the order of the rules is for this is largely due to the fact that the algorithm checks unimportant. The rules themselves are useful because the whole Moore neighborhood of a boundary pixel in they refer to variables and the adjectives that describe order to find the next boundary pixel. Unlike the Square those variables. Before you can build a system that Tracing algorithm, which makes either left or right turns interprets rules, you must define all the terms you plan on and misses "diagonal" pixels; Moore-Neighbor tracing using and the adjectives that describe them. will always be able to extract the outer boundary of any connected component. The reason for that is: for any 8- connected pattern, the next boundary pixel lies within the Moore neighborhood of the current boundary pixel. Since Moore-Neighbor tracing proceeds to check every pixel in the Moore neighborhood of the current boundary pixel, it is bound to detect the next boundary pixel. MATLAB shape measurement tool „Area‟ has been used for two consecutive images of input video frame to calculate change in the area. This is the scalar parameter and the actual number of pixels in the region. Fig. 6. General Fuzzy system. H. Neuro – Fuzzy algorithms Neural networks are composed of simple elements The fig (6) provides a roadmap for the fuzzy inference operating in parallel. These elements are inspired by process and it shows the general description of a fuzzy biological nervous systems. As in nature, the connections system. between elements largely determine the network function. You can train a neural network to perform a particular function by adjusting the values of the connections VI. IMPLEMENTATION (weights) between elements. Typically, neural networks Fig. 7 is the block diagram of the proposed system. are adjusted, or trained, so that a particular input leads to The implementation (except PLC & SCADA) has been a specific target output. Fig (5) illustrates such a situation. done MATLAB. There, the network is adjusted, based on a comparison of  CCD cameras are installed in the critical areas of the output and the target, until the network output the industry with proper scene planning. At single matches the target. Typically, many such input/target scene or single camera level, a multiple number of pairs are needed to train a network. fire suppression chemical control valves are Neural networks have been trained to perform complex installed which are connected to centralized functions in various fields, including pattern recognition, chemical pumping station shown in fig 8. identification, classification, speech, vision, and control systems. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 21 Automation by Using Neuro-Fuzzy Algorithms  Background image is stored in the reference  Fire area calculation is done by using region database and continuous video is acquired using properties. frame grabber.  Fire rules based on RGB, HSV, YCbCr color spaces  Median filter is applied for the reduction of salt and and change in the fire region area are applied. pepper noise.  Fire centroid pixel is identified.  Input RGB image and the Binary image are  If the existence of fire is confirmed, the image is multiplied. partitioned into four quadrants and the neural  Color segmentation is done. network is trained with fire quadrant pixel number  Fire area estimation is done. inputs.  Using neural networks computation of fire region is  Fire pixel edge detection and fire boundary done. estimation is done Fig. 7. Block Diagram. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 22 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms VII. ALGORITHMS & RESULTS Three color spaces are used to detect fire, to ensure high reliability. Change in fire area in two consecutive input video images is estimated in order to avoid false detection. The detailed system algorithm is as follows.  Read input image and background image - Fig 9. Fig. 8.(a) Fig. 9.(a) Fig. 9.(b) Fig. 9. (a) input img, (b) background img Fig. 8. (b)  Convert the RGB image into HSV equivalent and YCbCr equivalent. Also convert background RGB image into HSV image – Fig 10.  Continuously subtract the background image from input video frames - Fig 11.  Convert the remainder image to binary with fixed threshold and apply median filter - Fig 12. Fig. 8. (c) Fig. 8. (a) Camera installed, (b) Fire suppression control valves, Fig. 10.(a) (c) Centralized chemical/water pumping station.  Opening and closing parameters (values between 0 and 1) of fire control valve are calculated using fuzzy logic based on the fire region area and degree of brightness of the fire.  Valve control parameters are communicated via PROFIBUS to SCADA and PLC systems.  Fire suppression control valve supply (Current/ Voltage) is calculated in accordance with the fuzzy Fig. 10.(b) output using SCADA & PLC. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 23 Automation by Using Neuro-Fuzzy Algorithms Fig. 10.(c) Fig. 13. Image multiplication output Fig. 10. (a) & (b) HSV converted input and background images, (c) ycbcr converted input image.  Perform color segmentation and add all segmented images – Fig. 14. Fig. 14. (a) Fig. 11. Result of image subtraction Fig. 14. (b) Fig. 12.(a) Fig. 14. (c) Fig. 12.(b) Fig. 12. Binary image & filtered binary image Fig. 14. (d)  Multiply the input image and the binary image. Fig 13. Fig. 14. (a) Red, (b) orange, (c) Yellow, (d) addition of all color segments. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 24 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms  Extract RGB, HSV and YCbCr color spaces for applying of fire rules - Fig 15. Fig. 16.(b) Fig. 16. Edge and boundary of the image Fig. 15.(a)  Apply fire rules for detection of fire  Identify the fire region centre pixel number by using weighted centroid tool - Fig 17. Fig. 15.(b) Fig. 17. Blue dotted centroid pixels in fire centre  Partition the image with equal quantity of pixels in all regions and train the neural network – Fig 18. Fig. 15.(c) Fig. 15. (a) converted RGB image, (b) converted HSV image, (c) converted YCbCr.  Perform edge detection and boundary extraction - Fig. 18. (a): Four equal regions of image. Fig 16. Fig. 16.(a) Fig. 18. (b): NN training performance Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 25 Automation by Using Neuro-Fuzzy Algorithms Fig. 18. (c): NN training regression  Apply Fuzzy logic algorithm for opening and Fig. 19. (b): output based on input membership functions. closing of fire suppression control valve - Fig 19.  Convert 0 to 100% fuzzy output into 4-20 milliamps  Finally establish communication via PROFIBUS - to drive the control valve by using PLC & SCADA. Fig 20. Fig. 19. (a): Input parameter values Fig. 20. fire suppression control valve with 80% opening. Table 1. Results of automated fire suppression system. Cases Input image Output image Fire rules N-F parameter Automated action RGB √ RN 2 YCbCr √ ABV 252 Fire at Steel making furnace CA √ FAV 4300 service door. CVOP 60% RGB √ RN 3 YCbCr √ ABV 195 Fire on electrical CA √ FAV 2800 Cable trays. CVOP 50% RGB √ RN 4 Fire at YCbCr √ ABV 200 steel furnace CA √ FAV 1700 backflow discharge CVOP 50% Point. Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 26 Implementation of Computer Vision Based Industrial Fire Safety Automation by Using Neuro-Fuzzy Algorithms RGB √ YCbCr × Fire like CA × NO FIRE NO FIRE lamp CA: Change in area, RN: Region number, ABV: Average brightness value of region in 0-255, FAV: Fire area value in pixel numbers, CVOP: Control valve opening percentage VIII. DISCUSSION ON RESULTS REFERENCES 1 Tao Chen, Hongyong Yuan. An automatic fire searching Performance analysis has been carried out using five and suppression system for large spaces. Elsevier Fire sets of videos. Varieties of fire image frames and fire-like Safety Journal, 297-307, doi:10.1016/j.firesaf.2003. image frames are included. It has been observed that the 11.007, 2004. proposed system has very high fire detection and 2 MEI Zhibin, YU Chunyu. Machine Vision Based Fire automated fire suppression rate. Determination of change Flame Detection Using Multi-Features. 24th Chinese in fire area can differentiate most of fire-like regions with Control and Decision Conference (CCDC), 2844-2848, real fire regions. YCbCr color space has the ability of 2012. discriminating luminance from chrominance information. 3 A. Rehman, N. Masood. Autonomous Fire Extinguishing Since the chrominance dominantly represents System. IEEE Transactions, 218-222, 2012. 4 Hideaki Yamagishi. Fire Flame Detection Algorithm information without being affected by luminance, the Using a Color Camera. IEEE international symposium on chrominance-based rules and the color model defined in MHS,255-259, 1999. the chrominance plane are more reliable for fire detection. 5 Paulo Vinicius Koerich Borges. A Probabilistic Approach The basic challenge was the implementation of neuro- for Vision-Based Fire Detection in Videos. IEEE fuzzy algorithms for automated fire suppression system. transactions on circuits and systems for video technology, Trained neural network has given consolidated output vol. 20, no. 5, 721-731, 2010. based on targets to identify correct fire region in an 6 Bo-Ho Cho , Jong-Wook Bae. Image Processing-based image frame. Table 1 describes different sets of results, Fire Detection System using Statistic Color Model. which are obtained from critical industrial areas. In the International Conference on Advanced Language Processing and Web Information Technology, 245-250, first case steel making furnace service door has been 2008. considered. Due to improper sealing of furnace doors, 7 Mohammad Jane Alam Khan, Muhammed Rifat Imam. fire leakages occur. Another critical area is the electrical Automated Fire Fighting System with Smoke and cable system and which is explored in second case. Temperature Detection. IEEE International Conference on Overflow point in steel furnace is described in the third Electrical and Computer Engineering. 232-235, 2012. case where fire suppression agent is water. In the last a 8 Suzilawati Mohd Razmi, Nordin Saad. Vision-Based fire-like region was input to the system and it could Flame Detection: Motion Detection & Fire Analysis. IEEE identify that it is not fire. For the first three cases neural- Student Conference on Research and Development, 187- network is trained with as much as 19200 pixels for 191, 2010. 9 Andrey N. Pavlov, Evgeniy S. Povemov. Experimental single region and fuzzy algorithm is implemented with Installation for Test of Automatic Fire Gas Explosion suitable membership functions. All automated fire th Suppression System. 10 B IEEE International Conference suppression actions are shown in last column of table. and Seminar EDM, 332-334, 2009. 10 Turgay Celik, and Kai-Kuang Ma. Computer Vision Based Fire Detection in Color Images. IEEE Conference on Soft IX. CONCLUSIONS Computing in Industrial Applications, 258-263, 2008. 11 Jie Hou, Jiaru Qian, Zuozhou Zhao. Fire Detection Since fire can cause much damage and have Algorithms in Video Images for High and Large-span devastating consequences, great efforts are put into the Space Structures. IEEE Transactions, 2009. development of systems for its early detection and 12 KuoL. Su. Automatic Fire Detection System Using automated suppression. Video-based systems of fire Adaptive Fusion Algorithm for Fire Fighting Robot. IEEE detection have great advantages and can be used to International Conference on Systems Cybernetics, 966-971, overcome the shortcomings of sensor based systems In 2006. 13 Changwoo Ha, Ung Hwang. Vision-Based Fire Detection recent years, various fire detection systems using image Algorithm Using Optical Flow. IEEE Sixth International processing have been proposed, but many of them aim at Conference on Complex, Intelligent, and Software only fire detection. This paper has focused on automated Intensive Systems, 526-530, 2012. fire suppression based on fire video images. 14 Tian Qiu, Yong Yan. An Autoadaptive Edge-Detection Algorithm for Flame and Fire Image Processing. IEEE Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27 Implementation of Computer Vision Based Industrial Fire Safety 27 Automation by Using Neuro-Fuzzy Algorithms Transactions on Instrumentation and Measurement, 1486- PSIPL (ERM Group) as a project engineer in 2010 and started 1493, 2012. doing industrial research work as a Part time Master degree 15 Chen Jun, Du Yang, Wang Dong. An Early Fire Image research scholar. Published one international conference & Detection and Identification Algorithm Based on DFBIR three international Journal publications followed by one Model. IEEE World Congress on Computer Science and international journal book chapter in the field of computer Information Engineering, 229-232, 2009. vision based industrial automation. Presently working capacity 16 And& Neubauer. Genetic Algorithms in Automatic Fire as a Deputy Manager (Technical) at ERM Group steel & power Detection Technology. Genetic Algorithms in Engineering division. Systems: Innovations and Applications, 180-185, 1997. 17 T. Chen, P. Wu and Y. Chiou. An Early Fire-Detection Method Based on Image Processing. Proc. of IEEE Professor Dr. H.S.Mohana is born in the ICIP ‟04, 1707–1710, 2004. year 1965. Obtained B.E Degree in Electical 18 Lee and Dongil Han. Real-Time Fire Detection Using and Electronics Camera Sequence Image in Tunnel Environment. Engineering from University of Mysore Proceedings of ICIC, vol. 4681, 1209-1220, 2007. during 1986. Since then serving technical 19 Ishita Chakraborty, Ishita Chakraborty. A Hybrid education field in various capacities. Obtained Clustering Algorithm for Fire Detection in Video and M.E from University of Roorkee presently IIT Analysis with Color based Thresholding Method. IEEE ROORKEE with the specialization in International Conference on Advances in Computer Measurement and Instrumentation. Obtained Ph.D. form VTU Engineering. 277-280, 2010. during 2010 in Computer & Information Science. Worked as 20 Dengyi Zhang, Shizhong Han. Image Based Forest Fire chairmen and Member of Board of Examiner and Board of Detection Using Dynamic Characteristics With Artificial studies with several universities which includes, University of Neural Networks. IEEE International Joint Conference on Mysore, Kuvempu University and VTU. Presented research Artificial Intelligence, 290-293, 2009. findings in 12 National Conferences and in 4 International 21 Quanmin GUO Junjie DAI. Study on Fire Detection conferences held across the world. Recognized as AICTE Model Based on Fuzzy Neural Network. IEEE expert committee member in the inspection and reporting Transactions, 1-4, 2010. continuation of affiliation and Increase in intake of the 22 GUO Jian, ZHU Jie. Application of Self-Adaptive Neural Engineering Colleges. Completed, one AICTE/MHRD- Fuzzy Network in Early Detection of Conveyor Belt Fire. TAPTECH project, and one AICTE/MHRD- Research project IEEE Transactions, 978-983, 2009. successfully. Coordinated TWO ISTE Sponsored STTP for the 23 Turgay Çelik, Hü seyin Özkaramanl. Fire Pixel technical college teachers. Presently working as Dean Classification Using Fuzzy Logic and Statistical Color Academics in MCE Hassan, Karnataka, India. Model. IEEE Transactions, 1205-1208, 2007. 24 Ana Del Amo. Fuzzy Logic Applications to Fire Control Systems. IEEE International Conference on Fuzzy Systems, Professor Dr. P.A Vijaya obtained B.E 1298-1304, 2006. Degree in Electronics & Communication 25 Xuan Truong, Tung. Fire flame detection in video Engineering from University of mysuru, sequences using multi-stage pattern recognition techniques. Karnataka during 1985. Engineering Applications of Artificial Intelligence 1365- Since then serving technical education 1372, 2010. field in various capacities. Obtained M.E 26 Wirth,M. Zaremba,R. Flame region detection based on Degree from IISc Bangaluru during 1991 histogram backprojection. CRV 7th Canadian Conference and completed Ph.D in Pattern Recognition & Image on Computer and Robot Vision, 167-174, 2010. Processing during the year 2005 from IISc Bangaluru. Having 27 Xitao Zheng, Yongwei Zhang, Yehua Yu, Recognition of membership in professional bodies ISTE, IEEE & IACSIT and marrow cell images based in fuzzy clustering, published over 39 national & International journal publications. International Journal of Information Technology and Attended about 42 workshops and presented papers in various Computer Science (IJITCS), volume 1, Pages 40, 2012. countries across the world. 28 Hadi A. Alnabriss, Ibrahim S. I. Abuhaiba, Improved Image Retrieval with Color and Angle Representation, International Journal of Information Technology and Computer Science (IJITCS), Pp 68-81, 2014. 29 Amanpreet Singh, Preet Inder Singh, Prabhpreet Kaur, Digital Image Enhancement with Fuzzy Interface System, International Journal of Information Technology and Computer Science (IJITCS), Pages 51, 2012. Authors’ Profiles Mr. Manjunatha KC is born in the year 1985. Obtained B.E Degree in Instrumentation Engineering from Vesvesvaraya Technological University, Karnataka, India during 2008. Started career in teaching and served technical education field for 2 years. Joined steel manufacturing & power generation industry called M/s Copyright © 2015 MECS I.J. Information Technology and Computer Science, 2015, 04, 14-27

Advise: Why You Wasting Money in Costly SEO Tools, Use World's Best Free SEO Tool Ubersuggest.