TWI313848B - A fire image detection method - Google Patents

Description

1313848, ·9, invention description: [Technical field of invention] A method for detecting fire images, in particular, a similarity analysis between image pixels and a fire sample to know the similarity between image pixels and fire samples, To determine if there is a fire in the image. [Prior Art] In the technique of judging whether there is a fire after using the image to capture the surface, we can see the US patent 5,153,722, "pjre detecti〇n SyStem", which is first passed through two red/ultraviolet sensors. The illuminated area is sensed. If the detected energy exceeds the set threshold (Thresh〇ld), the image is captured by the camera, and the brightness area and boundary detection are analyzed. Edge detection) and image extraction (imagesubtmcti〇n) and other operations, to determine whether it is a real fire. Under this architecture, the detection of fire requires the use of an additional sensor to assist in the more accurate determination of the range of vibrations to be sensed. In addition, 'Wen-Bing Horng, JTian-Wen Peng, Chih-Yuan Chen 1313848, 2005, American Institute of Electrical and Electronics Engineers (IEEE, March 19-22, 2005) published a paper "A new image-based real In the -time flame detection method using color analysis", the detection process is divided into two parts: static and dynamic. The static samples are collected in advance, and the samples are collected from red, green, blue, and color spaces (Red, Green, Blue domain, RGB). Domain) is converted into HSI color space (HIS domain, Hue, Saturation, Luminance Intensity), and then the range of HSI values of sample fire and smoke is re-stated, and these values are used to identify fire and smoke. On the process. On the dynamic side, the image difference method is used to remove background objects similar to fire and smoke to determine whether there is fire and smoke. Second, at T. H_ Chen, P. H. Wu and Y. C. Chiou at the 2004 IEEE International Conference on Image Processing (International Conference on Image

Processing, ICIP, "An Early Fire-Detection Method Based on Image Processing" published in 2004) is divided into static and dynamic parts, while static parts are classified as fire. The features in the RGB color space (such as Rg 〇 > B), and the preliminary classification of the fire on the color of the 6 1313848, in terms of dynamics, use the image difference (] [11 ^ Ming difference) to remove the background similar to fire. The above two papers 'in terms of statics, are all using a large number of sources of wildfires'. The sub-systems analyze the characteristics or ranges of their specific color spaces (such as RGB &C; Cong) and have to pass multiple thresholds. In order to achieve the best results, go to the initial area of the fire. In this way, (4) the sample collection is time consuming and the calculation is complicated. SUMMARY OF THE INVENTION The present invention provides a method for detecting fire images, which uses a fire image sample as a basis, and the material sample of the age of the material is subjected to the classification method of the present invention, which can be (4) in the image of the simple image. Whether the pixel has a fire «, without the need for an additional device, does not require analysis of most fire image samples to solve the above problem. The fire image detecting method of the present invention is to fire a fire image sample with at least one captured image (the axis ge F is said to be miscellaneous, and (4) the image of the image of the image is broken - the fire pixel is not, the fire image _ The method package 7 1313848 includes: converting each sample pixel of the fire image sample into a color space, the color model having a first coordinate and a second coordinate, each of the sample pixels corresponding to the first coordinate and The second coordinate has a first coordinate value and a second coordinate value; calculating a center point of the first coordinate value and the second coordinate value of the sample pixels; and calculating each of the sample pixels of the fire image sample Obtaining a plurality of sample distance values from the Mahalanobis Distance of the center point; defining a threshold value according to the sample distance values; taking one of the plurality of image pixels of the captured image plane as the determination pixel; Calculating a Mahjong distance of the determination pixel from the center point to obtain a pixel distance; and comparing the pixel distance to the sample distance, and the pixel distance is 8 1313848 When the value is depreciated, it is determined that the captured image has the fire pixel. By the foregoing method, a fire image sample and the captured image can be used to determine whether there is a fire pixel to achieve the foregoing purpose. For a better understanding of the purpose of the present invention, the implementation method and the function thereof, the following is a detailed description of the following: [Embodiment] Please refer to "FIG. 1", which is a flowchart of the fire image detecting method of the present invention. The main purpose is to compare the fire image sample 30 (see "Fig. 2") with at least one captured image frame (40Γ image frame) to determine whether the captured image has a fire pixel; The fire image detection method includes: Step S12: converting each sample pixel of the fire image sample 30 into a color space, the color model having a first coordinate and a second coordinate, each sample pixel Corresponding to the first coordinate and the second coordinate having a first coordinate value and a second coordinate value; Step S14: calculating the first coordinate value and the second of the sample pixels a 9-1313848 one of the center points; a heart step S16: calculating a MahalanoWs Distance between the sample pixels of the fire image sample and the midpoint, and obtaining a plurality of sample distance values; Step S18: The sample distance value defines a threshold value. Step S20: taking one of the plurality of image pixels of the image plane 40 of the operation as a determination pixel; Step S22: calculating a Mahalanobis distance between the determination pixel and the center point to obtain a Pixel distance; and step S24: comparing the pixel distance to the sample distance, and determining that the captured image plane has the fire pixel when the pixel distance is less than the threshold value. Wherein, the color model in step S12 can be, for example, RGB, YIQ, HSV, HSI, etc., and the present invention is exemplified by a luminance color difference color model, wherein the first coordinate (Y) is brightness ( Luminance) coordinates, the second coordinate (1) is the color difference in the same direction (Inphase) coordinates, and the Q coordinate is the color difference quadrature component (Quadrat said) 'YIQ is the National Television System Committee (He added - 10 1313848

The Television System Committee (NTSC) stipulates that the following embodiments use the party coordinates and the color difference and the same component coordinates as an example, and the color model and the selection of the two coordinates are considered in various color models. In the luminance chromatic aberration color model, the luminance coordinate and the chrominance difference component coordinate are used in the practice of the present invention, and the threshold value is set relatively easily, and a relatively accurate judgment result can be obtained. The definitions of the color models listed above are briefly described below, in which the CMY (Cyan Magenta Yellow) color model is the most commonly used color model in the color image printing industry. In the color cube, the three colors are red and green. The blue complementary color, the CMY color model is the three colors to represent the various color models. The HSV (Hue, Saturation, Value) color model mainly corresponds to the color matching model of the family, which is consistent with the visual perception of the human eye. It is a model suitable for human eye resolution, in which Η defines the wavelength of color production, called color tone, S Defining the degree of lightness, called saturation, V defines the degree of lightness and darkness of the color, called brightness, usually expressed as a percentage. Secondly, in the method of converting sample pixels into a color difference color model, it is more common to convert the red and green 11 1313848 blue color model (RGB domain) into a luminance color difference color model (YIQ domain), so the following is listed. Formula:

Y 0.299 0.587 0.114 RI — 0.595716 -0.274453 -0.321263 GQ 0.211456 -0.522591 0.311135 B From the above formula, there is a high correlation between the YIQ color model and the RGB color model. The difference is that the YIQ coordinates represent brightness, The chromatic aberration is the same component and the color difference is orthogonal component, and RGB is the red, green and blue coordinates, respectively. After the sample pixels are converted, each sample pixel will have a first coordinate value and a second coordinate value corresponding to the Y, I coordinate, that is, the luminance value and the color value of the same component, of course, each sample pixel will also The color difference quadrature component value of the Q coordinate is used, but since this coordinate is not taken as an example in this embodiment, it will not be described again. Second, that is, the first coordinate value of the sample pixels and the center point of the second coordinate value are calculated in step S14, and the first coordinate value and the second coordinate value of all sample pixels are respectively calculated as arithmetic numbers thereof. The average number means that all the first coordinates 12 1313848 are added and divided by the number of sample pixels. All the second coordinate values are added and divided by the number of sample pixels, so that the coordinate value of the center point is obtained. The Markov distance of step S16 is a distance method in the cluster analysis in the statistical method, and the calculation formula is: M (Xz.) = -"Growing 1 " _ where A is (W X3,...,w , p is the coordinate value of the center point, Σ μ is the covariate matrix in the joint group, Dm (A) is the Markov distance of χ,·, and detailed description of the detailed calculation. This step is mainly to calculate the fire image. The Mahalanobis distance between each sample pixel and the center point in the sample to form a plurality of sample distances. The step of defining the threshold value in step S18 is based on the definition of the threshold value, which may be different depending on the distribution of the sample distance, and may have the following definitions. The method includes setting the threshold value to a value of the sample distance value greater than 95%, defining the threshold value as the median value of the sample distance values, setting the threshold value to the maximum value of the sample distance values, and setting the threshold 13 1313848 The value is set to the average of the sample distance values, the threshold value is set to the average of the sample distance values multiplied by a safety factor or the threshold value is set to the average of the sample distance values plus the samples Distance value The standard deviation is different, wherein the safety factor can be adjusted according to the actual situation, for example 12. The definition of various threshold values of Liszian changes depending on the histogram distribution of the sample distance, for example, if it is a normal distribution, it is three times the standard deviation. The method is preferably 'if the distribution pattern is known in the statistics', the maximum value, the sample distance value greater than a certain ratio (such as 95%), or the ratio of the fire pattern in the image sample is used to define the threshold value. 'The image surface of step S20 is the instant surface captured by the camera. The image surface is composed of a continuous frame, which is often used to make the image surface better in the human eye. For the effect, it is necessary to take out at least 30 frames in the second seconds; Step (10) The bribes of the towel drills can be taken in the age of the material, and one of them is the right hall in the image 40 If there is a shirt, the pixel is extracted as a broken pixel, and then the subsequent steps are carried out. In this way, more than 14 1313848 are required for the comparison of the pixel and the fire pixel, which will make the system load heavier, and may also be due to the performance of the system. Face The number of images is too large, so that the operation speed cannot keep up with the speed of each recording in the image plane. The second method can take a plurality of image pixels which are different from the previous image surface in the captured image frame. _ is the judgment pixel, which means that the image plane that is currently captured and the image acquired by the previous time point are image-by-aligned to find out the different image pixels, and then the phase The different shadow county money - set as _ material, (four) line subsequent steps. In step milk and step S24 is calculated the judgment pixel and the center point of the Markov distance (four) to get - pixel series, and _ the image is separated from the The distance between the sample and the image that is less than the distance of the pixel is the same as that of the pixel in the sample of the fine fire, so if the pixel is smaller than the threshold value, Dyna's "Riding Red _ like reading fire material samples belong to the same - cluster, that is, the judgment pixel should be a fire pixel. "For the example calculation of the present invention, please refer to "Fig. 2" and "3rd picture". "W" is the first example of the fire image sample. The figure is based on the fire image 15 1313848 sample. , this - "The image sample can be taken from any actual fire image. If the sample pixel of this fire image sample has 69*104 pixeis, that is, there are 7176 pixels, which are converted from RGB coordinates to γΐ(} coordinates. After that, the γ-seat and the I-score have a matrix of [69 x 104], and the two matrices are arranged in the following manner. Since the data is quite large, the present embodiment only proposes a specific numerical value as an explanation of the embodiment. : X. 7176 J 7176 , where 丫丨 is 79.1697, 1 丨 is 81.8977 as the following example. After step S14, the center point is obtained for her (10) by the cross-calculation to obtain 255.0835 and 40.8868 Υμ Ιμ, where Yu and The covariate matrix L -1 in the joint group of Iu Y and I is

7176 I

Mean\Xi - Meanf (7176)-1, and take X] as an example: 79.1697 81.8977 255.0835]^ 40.8868

79.1697 X 81.8977 '255.0835ί τ\ 40.8868 _ J 4 2.1291 -0.59841 = 104x -0.5984 0.1682 16 1313848 Then add the covariate matrix in the joint group one by one Σι+Σ2+...Σ and divide by (7176-1) to get 1.2086 -0.7754 = 103χ -0.7754 0.6762 where, in order to prevent the entire operation from being caused by dividing by zero,

The unit covariate matrix can also be obtained by adding a unit matrix to: 1.2096 -0.7754 103χ -0.7754 0.6772. The foregoing steps S16 and S18 are performed. In the embodiment of the present invention, the distribution of the sample distance values is not a normal allocation. And 95% of the sample distance value is less than 6, so the threshold value is set to 6. Thereafter, one of the plurality of image pixels of the captured image plane 40 (as shown in FIG. 3) is taken as the judgment pixel, and it is assumed that the 像素 and I coordinate values of the judgment pixel are 0, 0. The calculation of the Mahalanobis distance between the pixel and the center point is as follows:

[/ V - _ r, f \ a] ί _ T \ 0 225.0835 1.2096 -0.7754 \ 0 225.0835 D,= — X 103 XX — 1 } 0 40.8868 -0.7754 0.6772 0 40.8868 η ,匕” L - J )VL - J 17 1313848 D ^15 23 * 1 'Because its value is greater than the threshold value, it is not judged that the fire is a female pixel. For the decision to determine the pixel, please refer to "4th picture", which is another image of the image 41, which can be seen in the figure with the fire-like pixel 42' and in the image-only surface 40 There is no such flame pixel. Therefore, when judging the pixel image before and after the two images, the comparison is made, and the changed image pixel is calculated by the Mahalanobis distance, and the time of some operations can be omitted. After calculating the Mahalanobis distance, the flame pixel has a value of 3.25, which is smaller than the threshold value, that is, it is determined to be a fire pixel. Regarding the physical meaning of the foregoing method of the present invention, the cluster analysis (classificad〇n) in the statistical method is used to directly calculate the distance from each fire pixel to the center value in the fire image sample by the Mahalanobis distance method, After the experiment, the method of calculating the Markov distance is used to calculate the coordinates of the color space. It is necessary to match the luminance and chrominance components in the same direction, or the orthogonal components of the luminance and chromatic aberration, which will have a better effect. Therefore, at the threshold. After the value is defined, the judgment pixels in the newly captured image plane are calculated as the distance (4) of the distance between the 18 1313848 heart points in the sample, that is, the fire pixels in the sample fire image sample belong to the same cluster. To judge the fire pixel. «月> 考第5图' is a schematic diagram of a second embodiment of a fire image sample of the present invention, which takes (4) a sample sample as an example, generally in the case of a Zhao fire, usually accompanied by smoke, so smoke image The sample also needs to be used as an important basis for the judgment of the fire pixel. The image method and the above _ are not mentioned. The fire image samples 3〇, 31 can also be used as a fire image sample and a smoke image sample. Therefore, the center point includes a fire center point and a smoke center point, and the sample distance includes a fire sample distance and a The smoke sample distance, the pixel distance includes a fire pixel distance and a smoke pixel distance, the gate value includes a fire gate broadcast value and a smoke gate pinch value, when the fire pixel distance is less than the fire door broadcast value and the When the distance of the smoke pixel is less than the value of the smoke door, it is judged that the image of the captured image has a fire pixel, and thus the judgment is more rigorous, and it is determined that the fire pixel is generated by the fire and the smoke at the same time. The fire image sample 30 and the captured image surface 4, 41 are preferably implemented by the same image capturing system and the hardware to obtain a pixel of 19 1313848, in order to eliminate the hardware and hardware. The difference between the two. For the application of the present invention, please refer to "Fig. 6", which combines the foregoing steps S12, S14, S16, S18 into step S50: sample analysis to set the threshold value, and step S20, S22 is step S52, Calculating the pixel distance of the pixel, and then comparing the pixel distance with the threshold value in step S54, if the pixel distance is less than the threshold value, recording the image surface as a fire picture, and starting counting (S55), after the judgment is completed After an image is scanned, step S56 is performed to determine whether there are 30 consecutive images of the face of the image, and if so, an alarm is issued (S58). If not, step S59 is performed, If it is not a fire screen, the fire surface count will be zero. In this way, the alarm can be avoided. Of course, the design of the above 30 consecutive fire surface counts can be adjusted according to different needs. Twenty or more or less to meet actual needs. While the present invention has been described in its preferred embodiments, the present invention is not intended to limit the invention, and the invention may be modified and modified without departing from the spirit and scope of the invention. The scope of protection is defined by the scope of the patent application of PCT Application No. 20 1313848. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flow chart of a fire image detecting method according to the present invention; FIG. 2 is a schematic view showing a first embodiment of a fire image sample of the present invention; Schematic diagram of the captured image; • Figure 4 is a schematic diagram of another captured image; Figure 5 is a schematic view of a second embodiment of the fire image sample of the present invention; and Figure 6 Is a flowchart of an application example of the present invention. [Main component symbol description] 30,31 Fire image sample • 40,41 Image screen 42 Flame pixel step S12 converts each sample pixel of the fire image sample into a color space (Step S14) Calculate the center point of the sample pixel Step S16 Calculating the Mahalanobis distance between the sample pixel and the center point to obtain the sample 21 1313848. The distance value step S18 defines a threshold according to the sample distance value. Step S20 takes one of the image pixels of the image plane as the determination pixel. Step S22 calculates the judgment pixel and the center point. The pixel distance is obtained by comparing the pixel distance and the sample distance in step S24, and the image image has a fire pixel when the pixel distance is smaller than the sample distance. Step S50 Sample analysis step S52 Image capture step S54 Fire pixel determination step S55 fire After the face counting step S56, the fire surface count reaches 30 steps S58 is issued an alarm step S59. If the image surface is not a fire screen, the fire surface count is reset to zero.

Claims (1)

1313848 X. Patent application scope: 1 'A fire image_method, storage-fire image sample is compared with at least the image of the image (brieze frame) to determine that the image of the bed has a fire Pixel No, the fire image_method includes: the fire "the sample pixel of each sample sample - color space (Color Space)" the color model has a _ coordinate and a second coordinate of each sample pixel The ride should be the first - the second coordinate of the training has a first coordinate value and a second coordinate value; Calculating the first coordinate value of the sample pixels and the first center point; the coordinate value of the center point of the two coordinate values is a row break to calculate the Mahalanobis Distance of each of the sample pixels of the fire image sample And obtaining a plurality of deviation values; defining a threshold value according to the distance value of the temple sample. Taking 23 1313848 pixels of the plurality of image pixels of the captured image surface; calculating the Mahalanobis distance between the determination pixel and the center point a pixel distance; and a distance from the sample distance and the sample distance, and determining that the captured image has the fire pixel when the pixel distance is less than the threshold value. < 2. The fire image (4) method according to the scope of the patent scope, wherein the color model is a luminance color difference color model (YIQ d_in), and the first coordinate is a Luminance coordinate, the first The two coordinate systems are color--one-phase in-phase (Inphase) coordinates. - .. -....... _ 3. The method for detecting fire images according to claim 2, wherein the first coordinate value of the sample pixel is one The brightness value, the standard value of the sample pixel is a color difference same direction component value. 4. The method for detecting fire images as described in claim 1, wherein the Markov distance is (A.) = (χ. - μ)τ^ (χ. ~ , where a is ~ brother' The value and the two coordinate values of the brother, y is the central point of the 24 1313848 seat " ^ value and the second coordinate value, Σ 1 is the joint group covariate matrix is from (go) to \ the horse The method for detecting fire images as described in item 1 of the patent application scope, wherein the '^ gate detection value is greater than 95% of the sample distance values. Lu 6. As described in the scope of claim patent The method for manufacturing a fire image, wherein the threshold value is the median value of the sample distance values. 7. The fire image method according to claim W, wherein the value is the maximum value of the sample distance values. δ· The method for detecting fire images according to claim 1, wherein the root value is an average value of the distance values of the samples. 9. The fire image detection described in the scope of claim patent item In the test, the value of the door is multiplied by the average of the distance values of the samples, and a total factor of 6 females. The fire image pre-measurement method described in claim 9 wherein the safety factor is 1.2. 11. The fire image detecting party described in claim 1 of the patent application, wherein the sample is 25 1313848 The threshold value is the standard deviation of the average value of the sample distance plus three times the standard deviation. Among them 12. For the fire image detection method described in item 1 of the patent application, the plural of the captured image image is taken. One of the image pixels is a step of determining the pixel by taking one of a plurality of image pixels different from the front image edge in the captured image plane as the determination pixel. 13. For example, the patent application (4) The fire image_method, wherein the step of calculating a Mahjong distance between the determination pixel and the center point to obtain a pixel distance is to first convert the determination pixel into the first coordinate value and the second coordinate And then calculating the Mahalanobis distance of the judgment pixel and the center point. 14. The fire image method according to claim 1, wherein the fire image sample is a fire image sample. The fire image detecting method described in the scope of the patent application, wherein the image of the fire image is a smoke image sample. 16· The method for detecting fire images according to claim 1 of the patent scope, 26 1313848 The fire image sample is a fire image sample and a smoke image sample, the center point includes a fire center point and a smoke center point, and the sample distance includes a fire sample distance and a smoke sample distance, the pixel distance The method includes a fire pixel distance and a smoke pixel distance, the threshold value includes a fire door broadcast value and a smoke threshold value, and when the fire pixel distance is less than the fire door threshold value, and the smoke pixel distance is less than the smoke threshold value When judging, the captured image mask has a fire pixel. 17. The method for detecting fire images as described in claim 1 of the patent application further includes the step of s. a warning is issued when 30 consecutive images of the captured image have the fire pixels. The fire image forming method according to claim 1, wherein the step of determining the pixel distance from the center point to obtain a pixel distance further comprises converting the judgment pixel into the color model. The steps. The fire image detecting method according to claim 1, wherein the center point is an arithmetic mean of the first coordinate value of the first coordinate value of the sample pixels and an arithmetic mean of the second coordinate value. 28