JP6138036B2 - Flame detection device - Google Patents

Description

The present invention relates to a flame detection apparatus that detects a flame by performing image processing on an image captured by a surveillance camera, and more particularly to a flame detection apparatus that includes an algorithm for detecting a flame with high accuracy.

  The number of deaths due to fire has not been interrupted, and early detection of fire is very important from the viewpoint of initial fire extinguishment in the event of a fire or prevention of delay in a fire accident. Fires become large in a short time and damages become enormous. Therefore, it is especially important to suppress damage by early detection and extinguishing. Discovery by human visual inspection and smell is certain, but if it occurs in a place where it cannot be seen, it will be delayed and the labor force by monitoring will be large.

  Therefore, in the field of fire detection devices, research has been conducted on early detection of a fire factor by performing image processing on an image captured by a surveillance camera. Various methods for detecting a fire from a two-dimensional image have been proposed and some have been put to practical use. Fire detection targets are roughly classified into flames and smoke, and fire detection devices that directly detect the presence of flames have been put into practical use.

  As an example thereof, there is a conventional technique for reliably recognizing a fire flame by image processing without being affected by disturbance light such as illumination (for example, see Patent Document 1). The flame detection apparatus according to Patent Document 1 has a density distribution of a portion where a positional change is large for a specific color component in addition to a spatial color temperature detection process for detecting a region having a high color temperature based on image data of a warning area. A spatial frequency detection process for analyzing a certain spatial frequency is provided. By providing such a configuration, it is possible to reliably recognize a flame caused by a fire by distinguishing ambient light such as illumination, and to greatly improve the reliability of flame judgment by image processing.

Japanese Patent Laid-Open No. 10-126765

However, the prior art has the following problems.
Various factors other than ambient light such as illumination can be considered as factors that adversely affect the flame detection accuracy. Moreover, it is difficult for the technique of patent document 1 to detect the flame in a large-scale space at an early stage. Therefore, there is a demand for a flame detection apparatus that suppresses the influence of various disturbance factors and detects the early detection of a flame in a wide space with high accuracy using an image processing technique.

The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a flame detection apparatus capable of detecting a flame with high accuracy in various installation environments including various erroneous detection factors. And

  A flame detection apparatus according to the present invention is a flame detection apparatus that detects the occurrence of a flame by performing image processing on an image captured by a monitoring camera, and includes a plurality of images captured in time series by the monitoring camera. An image memory that stores images as time-series images, and a moving body that exists in the image based on a time-series image for one cycle that includes a plurality of N frames (N is an integer of 2 or more) stored in the image memory A flame detection unit that determines whether or not a flame has occurred by calculating a feature amount that is an index for determining whether or not a flame has occurred and performing a feature amount extraction process. A time difference image obtained as a difference between a time-series image and a reference image and a difference between adjacent images in the time-series image based on a time-series image for one cycle composed of a plurality of N frames. To the series difference image Then, a flame candidate area is extracted for each frame, each barycentric position of each flame candidate area extracted for each frame is calculated, a data sequence of barycentric positions for a plurality of N frames is subjected to discrete Fourier transform, and an average amplitude is calculated. A feature amount calculation unit that calculates a spectrum and a maximum amplitude spectrum, calculates a feature amount by the following formula, feature amount = maximum amplitude spectrum / average amplitude spectrum, and the feature amount calculated by the feature amount calculation unit is a predetermined value or more. In some cases, the flame candidate region has a flame determination unit that determines that it is not flame because of its strong periodicity.

The flame detection apparatus according to the present invention is a flame detection apparatus that detects the occurrence of a flame by performing image processing on an image captured by a monitoring camera, and is captured in time series by the monitoring camera. Based on an image memory that stores a plurality of images as time-series images and a plurality of N frames (N is an integer of 2 or more) stored in the image memory, the image exists in the image. A flame detection unit that extracts a moving object, calculates a feature amount that is an index for judging whether or not a flame has occurred, and performs feature amount extraction processing to judge whether or not a flame has occurred, and includes a flame detection The unit obtains a difference between a background difference image obtained as a difference between a time-series image and a reference image and a difference between adjacent images in the time-series image, based on a time-series image composed of a plurality of N frames. Time-series difference image Based on the above, a flame candidate area is extracted for each frame, the respective gravity center positions of the flame candidate areas extracted for each frame are calculated, and the increase / decrease number of the gravity center position between each adjacent frame in one cycle, Alternatively, when a feature amount calculation unit that calculates a feature amount based on a value obtained by performing discrete Fourier transform on a data sequence of centroid positions for a plurality of N frames, and the feature amount calculated by the feature amount calculation unit is greater than or equal to a predetermined value The flame candidate region has a flame determination unit that determines that it is not flame because of its strong periodicity.

According to the present invention, attention is paid to the rotating lamp that is one of the erroneous detection factors, flame determination is performed based on the feature amount indicating whether the moving body has periodicity due to rotation, and the rotating lamp and the flame are detected. By identifying the above, it is possible to obtain a flame detection device that can detect a flame with high accuracy in various installation environments including various erroneous detection factors.

It is a block diagram of the flame detection apparatus in Embodiment 1 of this invention. It is explanatory drawing regarding the extraction process of the moving body pixel by the pre-processing part of the flame detection apparatus in Embodiment 1 of this invention. It is explanatory drawing regarding the feature-value 1 extracted by the flame detection process in Embodiment 1 of this invention. It is explanatory drawing of the elliptical coincidence degree relevant to the feature-value 1 in Embodiment 1 of this invention. It is explanatory drawing regarding the correction procedure of a focus position by the feature-value calculation part in Embodiment 1 of this invention. It is explanatory drawing regarding the feature-value 2 extracted by the flame detection process in Embodiment 1 of this invention. It is explanatory drawing regarding the weak edge extracted by the feature-value calculation part in Embodiment 1 of this invention. It is explanatory drawing regarding the congestion degree of the weak edge calculated by the feature-value calculation part in Embodiment 1 of this invention. It is explanatory drawing regarding the feature-value 3 extracted by the flame detection process in Embodiment 1 of this invention. It is explanatory drawing regarding the feature-value 4 extracted by the flame detection process in Embodiment 1 of this invention. It is explanatory drawing regarding the feature-value 5 extracted by the flame detection process in Embodiment 1 of this invention.

Hereinafter, preferred embodiments of the flame detection device of the present invention will be described with reference to the drawings.
In the present invention, in a method of detecting a flame by detecting a fluctuation peculiar to a flame, in order to solve a problem that it was difficult to distinguish from a false alarm source showing a pixel change similar to a flame, such as flickering of illumination. The technical feature is to realize an improvement in flame detection accuracy by combining a region extraction method specialized for extraction of a flame and a feature amount extraction method focusing on a specific shape created by a flame.

Embodiment 1 FIG.
FIG. 1 is a configuration diagram of a flame detection apparatus according to Embodiment 1 of the present invention. The flame detection apparatus in the first embodiment includes an image memory 10, a preprocessing unit 20, and a flame detection unit 30. The image memory 10 is configured as an image memory for a plurality of frames so that images captured by the camera 1 can be stored as time-series data for a certain period in the past.

  The preprocessing unit 20 includes a moving body extraction unit 21 and a candidate area generation unit 22. Then, the pre-processing unit 20 determines pixels of a portion that seems to be a moving body based on the image of each frame included in the past fixed period (one cycle) captured by the camera 1 and stored in the image memory 10. It has a function of extracting and specifying as a candidate area for performing flame detection.

  The flame detection unit 30 includes a feature amount calculation unit 31 and a flame determination unit 32. Then, the flame detection unit 30 calculates a feature amount for detecting the presence / absence of a flame for the candidate area specified for each cycle by the preprocessing unit 20, and a flame is generated based on the calculation result. It has a function to determine whether or not.

Here, an outline of the detection principle of the present invention will be described.
First, pixels corresponding to the moving object are extracted from the image acquired from the camera 1 (step 1). Next, a flame candidate region is created based on the extracted pixels (step 2). Next, the feature quantity of the moving body extracted in the created flame candidate region is calculated (step 3). Finally, the flame is determined from the feature amount, and a fire is determined (step 4).

  Here, in the first embodiment, a case will be described as an example in which each frame process and each cycle process are separately performed and one cycle is processed as 64 frames. In the first embodiment, specifically, the process of steps 1 and 2 until the flame candidate area is created is repeated for one cycle as a frame process, and the flame candidate area extracted as a result of the one cycle process is extracted. In the next cycle, the feature amount calculation in step 3 and the fire determination in step 4 are performed.

  Then, the mobile body extraction process in step 1 is executed for each frame by the mobile body extraction unit 21 in the preprocessing unit 20, and the flame candidate area creation process in step 2 is performed in the candidate area generation unit 22 in the preprocessing unit 20. Each frame is executed, and at the end of each cycle, a candidate area (final candidate area) to be used in the subsequent flame detection unit 30 is specified. Further, the feature amount calculation process of step 3 is executed every cycle by the feature amount calculation unit 31 in the flame detection unit 30, and the fire determination process of step 4 is executed every cycle by the flame determination unit 32 in the flame detection unit 30. Is done. Therefore, description will be made below by dividing into processing executed by the preprocessing unit 20 every frame and processing executed by the flame detection unit 30 every cycle.

(1) Processing executed by the pre-processing unit 20 every frame The pre-processing unit 20 extracts a moving object from the camera video acquired by the camera 1. FIG. 2 is an explanatory diagram regarding extraction processing of a moving object pixel by the preprocessing unit 20 of the flame detection apparatus according to Embodiment 1 of the present invention. In FIG. 2, in order to make the explanation easy to understand, a case where the flame that is the original detection target is extracted as a moving object is illustrated in an illustration. Further, description will be made on the assumption that the moving body is brighter than the background.

FIG. 2A shows the following three images as images captured by the camera 1.
(Image a1) A reference image in a state where there is no moving object (corresponding to a background image where no flame exists), for example, corresponding to an image acquired in advance when the program is started.
(Image a2) Among images for one cycle acquired in time series, an image acquired at a certain time t−1 and an image in which a moving body is imaged on the left side is illustrated.
(Image a3) Among images for one cycle acquired in time series, an image acquired at a certain time t, and an image in a state in which the moving object existing on the left side at time t-1 has moved closer to the center Is illustrated.

Next, FIG. 2B shows the following two types of difference images obtained based on the three images of FIG.
(Image b1) A difference image between the image a2 and the image a3, and shows a time-series difference image. Specifically, based on the image a2 at time t-1 and the image a3 at time t, which are time series, for each corresponding pixel of both images, the absolute value of the difference between the luminance values (tone values) of both pixels is calculated. The image generated by obtaining is referred to as a time-series difference image.
(Image b2) A difference image between the image a1 and the image a3, and shows a background difference image. Specifically, based on the reference image a1 and the image a3 at time t, for each corresponding pixel of both images, an image generated by obtaining the absolute value of the difference between the luminance values of both pixels is represented as a background difference image. It is called.

Next, FIG. 2C shows the following two binary images obtained by binarizing each of the two types of difference images in FIG. 2B with a predetermined threshold. Has been.
(Image c1) A binary image obtained by binarizing the image b1 with a predetermined threshold, and pixels having a difference equal to or larger than the predetermined threshold are extracted as a time-series difference image. Therefore, in the case of shaking like a flame, it is possible to extract pixels detected as a flame only at one time in the images at time t-1 and time t.
(Image c2) A binary image obtained by binarizing the image b2 with a predetermined threshold, and pixels having a difference equal to or larger than the predetermined threshold are extracted as the background difference image. Therefore, the pixel detected as flame can be extracted from the image at time t.

Next, FIG. 2D shows the next binary image obtained by calculating the logical sum of the image c1 and the image c2.
(Image d1) Based on the binary image c1 based on the time-series difference and the binary image c2 based on the background difference, black pixels are black in at least one image, and white pixels are white in both images. Thus, the flame detection area can be extracted.

  It is assumed that the pixels detected as flames have a predetermined high luminance value or higher. Therefore, when calculating the time-series difference image b1 or the background difference image b2 based on the two images, only a pixel having a predetermined high luminance value or more is extracted from at least one image to generate a difference image. Thus, it is possible to prevent the extra area from being finally extracted as the flame detection area.

  Then, the moving body extraction unit 21 in the preprocessing unit 20 repeatedly executes the series of processes shown in FIGS. 2A to 2C for each frame. In addition, the candidate region generation unit 22 in the preprocessing unit 20 finally performs a logical sum of one cycle (64 frames) on the flame detection region extracted for each frame as shown in FIG. The flame candidate area after one cycle processing is specified. That is, the candidate area generation unit 22 adds up pixels extracted as flame candidate areas even once in 64 frames, and identifies flame candidate areas.

  Further, the candidate region generation unit 22 counts the number of extractions for each pixel by counting the number of times extracted as a flame candidate region (hereinafter, this number is referred to as the number of extractions) within one cycle for each pixel. Can be calculated in advance. That is, the number of extractions is an index indicating how many of the 64 frames correspond to the pixels identified as flame candidate regions.

  The candidate area generation unit 22 can also exclude pixels from which the number of extractions is less than a predetermined number from the candidate area.

  In addition, the candidate area generation unit 22 sets a circumscribed rectangle for a set of pixels that form an island adjacent to each other among the pixels extracted as a flame candidate area even once in 64 frames. Thus, the flame candidate region can be specified as a rectangular region.

(2) Processing executed in each cycle by the flame detection unit 30 The feature amount calculation in the flame detection unit 30 in the first embodiment is performed for the flame candidate region specified for each cycle processing by the preprocessing unit 20. The unit 31 performs flame discrimination processing based on the following five feature amounts.
[Flame discrimination processing 1] The ellipse coincidence rate is extracted as the feature quantity 1, and flame discrimination is performed.
[Flame discrimination processing 2] The ratio of weak edges is extracted as the feature amount 2 and flame discrimination is performed.
[Flame discrimination processing 3] The periodicity of the transition of the center of gravity is extracted as the feature amount 3, and flame discrimination is performed.
[Flame discrimination processing 4] The increase / decrease state of the transition of the center of gravity position is extracted as the feature amount 4, and flame discrimination is performed.
[Flame discrimination processing 5] The change amount of the average luminance is extracted as the feature amount 5, and flame discrimination is performed.

  And the flame determination part 32 in the flame detection part 30 in this Embodiment 1 integrates the result of the flame discrimination | determination processes 1-5 by the feature-value calculation part 31, and finally performs a flame determination. Therefore, next, the details of each of the flame determination processes 1 to 5 will be described.

[Flame Discrimination Processing 1] Method of Extracting Ellipse Matching Ratio as Feature Quantity 1 and Performing Flame Discrimination The flame to be detected is different from other moving bodies, for example, the movement of a person, and has a shape close to an elliptical shape. It tends to be recognized. Therefore, in the flame discrimination process 1, an ellipse corresponding to the pixel data of the moving object existing in the identified candidate area is extracted, and the pixel identified as the flame candidate area is extracted from the extracted oval area. The contained ratio is calculated as an elliptical coincidence rate, and when the size of the elliptical coincidence rate is equal to or greater than a predetermined coincidence rate, it is determined that the detected object is a flame. Therefore, a specific determination method will be described below.

  An ellipse is defined as a “curve formed from a set of points such that the sum of distances from two fixed points is constant”. Therefore, the feature amount calculation unit 31 in the flame detection unit 30 extracts an elliptical measurement region in the following procedure based on the flame candidate region specified by the candidate region generation unit 22.

(Procedure 1) The feature amount calculation unit 31 extracts a moving body in each frame in the current cycle in the flame candidate region specified in the previous cycle. The extraction of the moving body is performed based on the time-series difference image and the background difference image in the same manner as the candidate area is specified by the preprocessing unit 20. Further, for each pixel in the flame candidate region, how many frames are extracted as a moving body in one cycle is counted as the number of extractions. Then, the feature amount calculation unit 31 performs gradation division according to the number of extractions. For example, the number of extractions that are counted with the minimum being 0 and the maximum being 64 may be divided into, for example, the following 8 gradations.
1st gradation: Extraction times 0-8
Second gradation: Extraction frequency 9-16
:
8th gradation: number of extractions 57-64

FIG. 3 is an explanatory diagram related to the feature quantity 1 extracted in the flame detection process according to Embodiment 1 of the present invention. FIG. 3A shows a specific example in which the result of gradation division is displayed by color.
Note that the above-described gradation division is merely an example, and the gradation can be divided into an arbitrary number of gradations, and the number of extractions included in each gradation can be set within an arbitrary range.

(Procedure 2) The feature quantity calculation unit 31 has a region with the maximum number of extractions (corresponding to a region identified as the innermost color in FIG. 3A) and a region with the minimum number of extractions (FIG. 3 ( (corresponding to the area specified as the outermost color in a)), the center of gravity is obtained using the following equation (1). Specifically, for each region, the position of the center of gravity obtained by calculating the average of the X coordinate and the Y coordinate for the pixels in the region is used as the focal point coordinate.

(Procedure 3) The feature amount calculation unit 31 sets the focus calculated for the region where the number of extractions is the maximum as the focus A and the focus calculated for the region where the number of the extractions is the minimum as the focus B. FIG. 3B illustrates the calculated positions of the focal point A and the focal point B. Then, the feature amount calculation unit 31 obtains the average of the sum of the distances from the focal point A and the focal point B by using the following expression (2) for the pixels included in the region where the number of extractions is the smallest, and the radius of the ellipse And

The variables in the above formulas (1) and (2) have the following contents, respectively.
F_x: X coordinate of the centroid position F_y: Y coordinate of the centroid position Px: X coordinate of the pixel included in the area Py: y coordinate of the pixel included in the area Pn: Total number of pixels included in the area FAx: X coordinate of the focal point A FAy: Y coordinate of focus A FBx: X coordinate of focus B FBy: Y coordinate of focus B i: Variable of X coordinate j: Variable of Y coordinate

  The flame burns violently while being shaped like an ellipse. However, when a person moves, the flame tends to have a complicated shape different from the elliptical shape caused by the flame. Therefore, the feature amount calculation unit 31 determines whether a pixel having the number of extraction times of 1 or more in the flame candidate region is present within or outside the ellipse defined by the above formulas (1) and (2). To do. Then, the degree of coincidence of the ellipses is calculated using the following formula (3).

  Then, the feature amount calculation unit 31 determines that the moving body detected in the flame detection area is a flame when the degree of coincidence obtained by the above equation (3) is equal to or greater than a predetermined allowable coincidence. can do. FIG. 4 is an explanatory diagram of the degree of oval coincidence related to the feature quantity 1 in the first embodiment of the present invention. A black area in FIG. 4A indicates an inner area of an ellipse defined by the above equations (1) and (2) among the flame candidate areas. In addition, a black area in FIG. 4B indicates an outer area of the ellipse defined by the above formulas (1) and (2) among the flame candidate areas.

  FIG. 4C shows the transition of the degree of coincidence calculated in each cycle when the flame is generated and the degree of coincidence calculated in each cycle when the person is moving. Comparison with transition is shown. As can be seen from FIG. 4C, by focusing on the degree of coincidence of the ellipses as the feature amount 1, it is possible to identify a flame and a moving body other than a flame such as a movement of a person.

  In the focus calculation based on FIG. 4 described above, the center of gravity is obtained for the region where the number of extractions of the changed pixels is minimum and maximum, and the elliptical region and the moving object are created by using these two points as the focal point. The degree of coincidence with the extracted area is determined as the feature amount 1. However, depending on the shape of the flame, this method may not be able to properly create an elliptical region.

  Therefore, in such a case, it is conceivable to improve the creation of the elliptical region by correcting the focus. FIG. 5 is an explanatory diagram regarding a focal position correction procedure by the feature amount calculation unit 31 according to the first embodiment of the present invention. The focal position correction procedure will be described with reference to FIG.

(Correction Procedure 1) First, the feature quantity calculation unit 31 rotates the major axis of the ellipse by a predetermined angle (for example, 10 degrees) clockwise and counterclockwise with respect to the middle point of the ellipse (FIG. 5A). reference).
(Correction Procedure 2) Next, the feature quantity calculation unit 31 limits the pixels included in the area surrounded by the two rotated straight lines (the area sandwiched between the two straight lines) to the above formula (1 ) To find a new center of gravity. In that case, a larger weight is added as the distance from the center of the ellipse becomes larger (see FIG. 5B).
(Correction Procedure 3) Finally, the feature quantity calculation unit 31 newly finds two newly obtained centroids by using the above equation (2) for the pixels included in the region where the number of extractions is the smallest. The average of the sum of the distances from the focal point A and the focal point B is obtained and set as the ellipse radius (see FIG. 5C).

  By executing such correction procedures 1 to 3, the focal point A and the focal point B are set on the outer side, so that a larger part of the flame can be accommodated in the newly defined ellipse. As a result, the degree of coincidence according to the above equation (3) can be increased, and the discrimination accuracy based on the feature amount 1 can be improved.

[Flame Discriminating Process 2] Method for Extracting the Ratio of Weak Edges as Feature Quantity 2 and Performing Flame Discrimination FIG. 6 is an explanatory diagram relating to the feature quantity 2 extracted in the flame detecting process according to Embodiment 1 of the present invention. is there. FIGS. 6A and 6B in the upper stage show an original image and an edge detection result when a flame is imaged. Further, FIGS. 6C and 6D in the lower stage show an original image and an edge detection result when a person's movement that is a false detection factor is imaged.

  As shown in FIGS. 6 (a) and 6 (b), the flame burns while emitting intense light, so that the high-intensity areas are dense, and as a result, the weak edges tend to be dense inside the flame. On the other hand, as shown in FIGS. 6C and 6D, such a tendency is not observed in the movement of the person.

  Therefore, in the flame discrimination process 2, the density of the edges of the moving object existing in the specified candidate area is calculated, and when it is composed of weak edges of a predetermined value or less, what is detected as the moving object is the flame. It is determined that Therefore, a specific determination method will be described below.

(Procedure 1) The feature quantity calculation unit 31 first extracts weak edges in order to obtain the density of weak edges. In the first embodiment, as an example, an edge having an edge strength of 15 or less is set as a weak edge. FIG. 7 is an explanatory diagram relating to weak edges extracted by the feature amount calculation unit 31 according to Embodiment 1 of the present invention. Specifically, the result of extracting the weak edge in the flame candidate region with respect to the edge detection result of FIG. 7A is shown in FIG. 7B.

(Procedure 2) Subsequently, the feature quantity calculation unit 31 obtains the density of pixels extracted as weak edges. First, the feature amount calculation unit 31 performs raster scanning on the entire image. If there is a pixel with a weak edge extracted, the feature amount calculation unit 31 sets a 5 × 5 small region around the pixel and exists within the small region. The number of pixels extracted as a weak edge is obtained. The size of the small area is not limited to 5 × 5, and a small area of (2m + 1) × (2m + 1) (m is an integer of 1 or more) can be set around the weak edge pixel.

  FIG. 8 is an explanatory diagram regarding the density of weak edges calculated by the feature amount calculation unit 31 according to Embodiment 1 of the present invention. FIG. 8A shows a weak edge detection result similar to FIG. 7B. FIG. 8B shows an enlarged view including a 5 × 5 small region in the upper left among the regions detected as weak edges in FIG. FIG. 8B illustrates a case where 12 pixels extracted as weak edges are included in a 5 × 5 small region.

(Procedure 3) Finally, the feature quantity calculation unit 31 obtains the number of pixels extracted as weak edges existing in the small area for all weak edge pixels, and calculates the average value as the density of weak edges. And And the feature-value calculation part 31 can judge that the moving body detected within the flame detection area is a flame when the density is a predetermined threshold value (for example, 15 pixels) or more.

[Flame discrimination processing 3] About the method of extracting the periodicity of the transition of the center of gravity position as the feature amount 3 and performing the flame discrimination In this flame discrimination processing 3, when the rotating lamp that is a false detection factor is extracted as a moving object, A method for distinguishing from flame will be described. When the extracted moving body is a rotating lamp, the position of the center of gravity of the moving body calculated for each frame has a transition with a strong periodicity as compared with the flame. FIG. 9 is an explanatory diagram related to the feature amount 3 extracted in the flame detection process according to Embodiment 1 of the present invention.

  Specifically, FIG. 9A shows the transition of the center of gravity calculated for each frame with respect to the moving body extracted by capturing a scene in which three revolving lights blink, in the X and Y coordinates. They are shown separately. FIG. 9 (b) shows the transition of the center of gravity calculated for each frame separately for the moving object extracted by imaging a scene in which a general flame burns, divided into an X coordinate and a Y coordinate. It is.

  FIG. 9C shows the result of performing discrete Fourier transform on the transition of the center of gravity position in FIG. 9A when the rotating lamp is extracted as a moving object, and is divided into X and Y coordinates. It is shown. Further, FIG. 9 (d) shows the result of performing discrete Fourier transform on the transition of the center of gravity position of FIG. 9 (b) when the flame is extracted as a moving body, divided into X and Y coordinates. It is a thing.

  In the flame discrimination process 3, discrete Fourier transform is used to quantitatively evaluate the periodicity of the center of gravity transition. Discrete Fourier transform is the decomposition of a waveform into a clean collection of sine waves. If the input waveform is a beautiful waveform, a strong spectrum can be obtained only in a certain frequency band. Therefore, a specific determination method will be described below.

(Procedure 1) The feature amount calculation unit 31 uses the above equation (1) to calculate the position of the center of gravity for each moving candidate extracted for each frame of the current cycle for each flame candidate region identified in the previous cycle. Is calculated. The transition state of the center of gravity based on the result corresponds to FIGS.
(Procedure 2) Next, the feature quantity calculation unit 31 performs a discrete Fourier transform on the transition data of the centroid position obtained in the procedure 1. The result of showing the relationship between the frequency and the amplitude spectrum in a graph corresponds to FIGS. 9C and 9D. In FIGS. 9C and 9D, as an example, discrete Fourier transform is performed on data for 64 frames (one cycle).

(Procedure 3) Finally, the feature quantity calculator 31 obtains an average amplitude spectrum and a maximum amplitude spectrum based on the result of the discrete Fourier transform in Procedure 2. Furthermore, the feature amount calculation unit 31 calculates, as the feature amount 3, how large the maximum spectrum is with respect to other spectra by using the following equation (4).

  It is considered that the periodicity increases as the value of the feature amount 3 increases. Therefore, the feature amount calculation unit 31 determines that the periodicity is strong when the feature amount 3 becomes a value equal to or greater than a certain value, and the moving body detected in the flame detection region is not a flame but a rotating lamp with a false detection factor. It can be judged that there is, and false alarms can be suppressed.

[Flame discrimination processing 4] About the method of extracting the increase / decrease state of the center of gravity position as the feature amount 4 and performing the flame discrimination In this flame discrimination processing 4, the person or the vehicle that is a false detection factor is extracted as the moving body. In this case, a method for distinguishing from flame will be described. When the extracted moving body is a person or a car, these moving bodies often move straight, and the transition of the center of gravity tends to increase or decrease monotonously. On the other hand, when the extracted moving body is a flame, the transition of the center of gravity tends to increase or decrease as the flame fluctuates. FIG. 10 is an explanatory diagram related to the feature amount 4 extracted in the flame detection process according to Embodiment 1 of the present invention.

  Specifically, FIG. 10 (a) shows the transition of the center of gravity calculated for each frame for a moving body extracted by imaging a scene in which a general flame burns, divided into an X coordinate and a Y coordinate. It is shown. FIG. 10B shows the transition of the gravity center position calculated for each frame separately for the moving object extracted by capturing the scene where the person moves, divided into the X coordinate and the Y coordinate. . A specific determination method by the feature amount calculation unit 31 will be described below.

(Procedure 1) The feature amount calculation unit 31 uses the above equation (1) to calculate the position of the center of gravity for each moving candidate extracted for each frame of the current cycle for each flame candidate region identified in the previous cycle. Is calculated. The transition state of the center of gravity based on the result corresponds to FIGS.
(Procedure 2) Next, the feature quantity calculator 31 calculates the feature quantity 4 according to the following equation (5) based on the transition data of the center of gravity obtained in the procedure 1.

  If the value of the feature amount 4 increases, it is considered that the tendency of monotonic increase or monotonic decrease is strong. Therefore, the feature amount calculation unit 31 determines that the position of the center of gravity is monotonously increasing or monotonically decreasing when the feature amount 4 is a certain value or more, and the moving object detected in the flame detection region is not a flame. Therefore, it can be determined that the object is a straight-moving body such as a person or a vehicle that is an erroneous detection factor, and erroneous reporting can be suppressed.

[Flame Discriminating Process 5] A Method for Extracting the Change in Average Brightness as the Feature Quantity 5 and Performing Flame Discrimination In this flame discriminating process 5, when a person or a vehicle that is a false detection factor is extracted as a moving object, A method different from the flame discrimination process 4 will be described. When the extracted moving body is a person, a car, or the like and stays on the spot, the change in luminance value becomes gentle because there is not much change in pixels. On the other hand, when the extracted moving body is a flame, the luminance change tends to increase due to the image of the fiery burning state. FIG. 11 is an explanatory diagram related to the feature quantity 5 extracted in the flame detection process according to Embodiment 1 of the present invention.

  Specifically, FIG. 11A shows the transition of the average brightness value calculated for each frame for a moving object extracted by imaging a scene in which a general flame burns. FIG. 10B shows the transition of the average luminance value calculated for each frame with respect to a moving object extracted by capturing a scene in which a person moves. A specific determination method by the feature amount calculation unit 31 will be described below.

(Procedure 1) The feature amount calculation unit 31 uses the following equation (6) to perform one cycle for each moving body extracted for each frame of the current cycle for each flame candidate region identified in the previous cycle. The sum of the luminance value change amounts at is obtained as the feature amount 5.

(Procedure 2) Next, when the feature quantity 5 obtained in the procedure 1 is equal to or less than a predetermined quantity, the feature quantity calculation unit 31 has little change in the average luminance value of each frame calculated in one cycle. Therefore, it can be determined that the moving object detected in the flame detection area is not a flame but is in a state where a false detection factor such as a person or a vehicle stays, and erroneous reporting can be suppressed.

  As described above, according to the first embodiment, an appropriate candidate region for flame detection can be specified for each cycle based on the latest time-series image data by the action of the preprocessing unit. Furthermore, for the candidate area identified in the previous cycle, by performing determination processing based on the feature amount 1 to the feature amount 5 in the next cycle, it is possible to suppress false detection and improve the detection accuracy of the flame itself. Can do. As a result, it is possible to realize a flame detection apparatus and a flame detection method that can detect a flame with high accuracy in various installation environments including various erroneous detection factors.

  In addition, although the flame determination part 32 in the flame detection part 30 of this invention integrates the result of the flame determination processes 1-5 by the feature-value calculation part 31, and finally performs the flame determination, The sensitivity may be changed by weighting each feature amount, and more specifically, when at least one of the feature amounts exceeds a predetermined threshold, it may be determined that the flame is present.

  DESCRIPTION OF SYMBOLS 1 Camera, 10 image memory, 20 Pre-processing part, 21 Mobile body extraction part, 22 Candidate area | region production | generation part, 30 Flame detection part, 31 Feature-value calculation part, 32 Flame determination part

Claims (3)

A flame detection device that detects the occurrence of a flame by performing image processing on an image captured by a monitoring camera,
An image memory for storing a plurality of images taken in time series by the monitoring camera as time series images;
Based on a one-cycle time-series image composed of a plurality of N frames (N is an integer of 2 or more) stored in the image memory, a moving object existing in the image is extracted to determine whether or not a flame has occurred. A flame detection unit that determines whether or not a flame has occurred by calculating a feature quantity serving as an index to perform and performing feature quantity extraction processing,
The flame detection unit is based on a time-series image for one cycle composed of the plurality of N frames.
Flame candidates for each frame based on a background difference image obtained as a difference between the time-series image and a reference image and a time-series difference image obtained as a difference between adjacent images in the time-series image. A region is extracted, and each centroid position of the flame candidate region extracted for each frame is calculated. A data sequence of the centroid positions for the N frames is subjected to discrete Fourier transform to calculate an average amplitude spectrum and a maximum amplitude spectrum. And the following formula: feature quantity = maximum amplitude spectrum / feature quantity calculation unit for calculating the feature quantity by the average amplitude spectrum;
A flame detection device comprising: a flame determination unit that determines that the flame candidate region is not a flame because the periodicity is strong when the feature amount calculated by the feature amount calculation unit is greater than or equal to a predetermined value. The flame detection apparatus according to claim 1,
The feature amount calculation unit includes:
(Process 1) The background is obtained by calculating the absolute value of the corresponding pixel difference for the nth image (n is an integer not less than 2 and not more than N) of the reference image and the time-series images for one cycle. Generate a difference image,
(Process 2) The time-series difference image is generated by calculating the absolute value of the difference between corresponding pixels of the n-th image and the (n-1) -th image among the one-cycle time-series images. ,
(Process 3) A set of pixels having a difference value equal to or greater than the binarization threshold by binarizing each of the background difference image and the time-series difference image based on a predetermined binarization threshold. A background difference binary image and a time series difference binary image configured as
(Process 4) The flame candidate region is extracted by taking a logical sum of the background difference binary image and the time-series difference binary image,
(Process 5) By executing Process 1 to Process 4 by sequentially changing n from 2 to N, the extraction of the flame candidate region is performed (N-1) times, and (N-1) times of the flames are performed. Calculate the center of gravity of each candidate area,
(Process 6) Discrete Fourier transform is performed on the data sequence of the centroid positions for the plurality of N frames to calculate the average amplitude spectrum and the maximum amplitude spectrum, and the feature amount is calculated by the following formula: feature amount = maximum amplitude spectrum / average amplitude spectrum A flame detection device. A flame detection device that detects the occurrence of a flame by performing image processing on an image captured by a monitoring camera,
An image memory for storing a plurality of images taken in time series by the monitoring camera as time series images;
Based on a one-cycle time-series image composed of a plurality of N frames (N is an integer of 2 or more) stored in the image memory, a moving object existing in the image is extracted to determine whether or not a flame has occurred. A flame detection unit that determines whether or not a flame has occurred by calculating a feature quantity serving as an index to perform and performing feature quantity extraction processing,
The flame detection unit is based on a time-series image for one cycle composed of the plurality of N frames.
Flame candidates for each frame based on a background difference image obtained as a difference between the time-series image and a reference image and a time-series difference image obtained as a difference between adjacent images in the time-series image. A region is extracted, and the centroid position of each of the flame candidate regions extracted for each frame is calculated. The number of centroid positions between adjacent frames in the one cycle, or the centroid positions for the N frames. A feature amount calculation unit that calculates a feature amount based on a value obtained by performing a discrete Fourier transform on the data string ;
A flame detection device comprising: a flame determination unit that determines that the flame candidate region is not a flame because the periodicity is strong when the feature amount calculated by the feature amount calculation unit is greater than or equal to a predetermined value.

Images