WO2014000427A1

WO2014000427A1 - Method and device for detecting flame

Info

Publication number: WO2014000427A1
Application number: PCT/CN2013/070026
Authority: WO
Inventors: 占伟鹏; 张翼; 叶茂; 庞锋
Original assignee: 华为技术有限公司
Priority date: 2012-06-29
Filing date: 2013-01-04
Publication date: 2014-01-03
Also published as: CN103514430B; CN103514430A

Abstract

Disclosed are a method and a device for detecting a flame. The method comprises: determining a candidate flame area in a current frame image; performing real-time luminous flux calculation on the candidate flame area based on a grid algorithm, to obtain a luminous flux vector of a pixel point of the candidate flame area; and determining whether a flame exists according to distribution characteristics of the luminous flux vector of the pixel point of the candidate flame area in each direction. The method and the device for detecting a flame of embodiments of the present invention are capable of accurately and stably detecting a flame in various complex or simple environments.

Description

Method and device for detecting flame

The present application claims priority to Chinese Patent Application No. 201210219956.1, entitled "Method and Apparatus for Detecting Flames", filed on Jun. 29, 2012, the entire contents of which is incorporated herein by reference.

Technical field

The present invention relates to the field of information technology and, more particularly, to a method and apparatus for detecting a flame. Background technique

A flame is a visible light or other physical manifestation that occurs during the chemical process in which a fuel and air are mixed and rapidly converted into a combustion product. Flames can bring many benefits, but they can be harmful if they are used carelessly. The most common is fire. Fires have always been a huge threat to the safety of people's lives and property. Being able to alert sudden fires in a timely manner is an effective way to reduce or avoid the damage caused by fire. In recent years, with the rapid development of computer vision technology, it has become possible to use this technology for early warning of fire, and some existing technologies have been applied. These technologies have shown good results in some indoor stable conditions, such as factories, hotels, houses, etc.

The prior art mainly focuses on analyzing the color characteristics and the motion characteristics of the flame. The color feature analysis of the flame is to divide the color image collected by the monitoring device into red, green and blue RGB channels, and set some of these three channels for each pixel. The threshold condition is such that the pixel points satisfying these relationships are discriminated as flame pixel points, and constitute candidate regions, among which there are three primary color component difference methods, dynamic threshold methods, and the like. In the analysis of the flame motion characteristics, the target object is separated from the background by the motion characteristics of the target object, and the static object with flame color characteristics is excluded, and then geometric appearance, irregularity, flicker frequency and the like are calculated, one of which is Or a plurality of features for comprehensive judgment to achieve the purpose of flame detection in the video. However, for the geometric appearance and irregularity of the flame, due to the open environment in some open scenes: wilderness, forests, streets, etc. (for example, strong winds appearing in open air scenes), the growth characteristics of the flame It cannot be judged only by the geometric shape, irregularity, etc. on the two-dimensional image. Therefore, the accuracy of existing flame detection methods cannot be guaranteed.

Summary of the invention

Embodiments of the present invention provide a method and apparatus for detecting a flame, which are capable of accurately and stably detecting a flame.

In one aspect, an embodiment of the present invention provides a method of detecting a flame, the method comprising: determining when a candidate flame region in the front frame image; performing real-time optical flow calculation on the candidate flame region based on a grid algorithm to obtain an optical flow vector of the pixel of the candidate flame region; and according to the optical flow vector of the pixel of the candidate flame region The distribution characteristics in all directions to determine if there is a flame.

In another aspect, an embodiment of the present invention provides an apparatus for detecting a flame, the apparatus comprising: a determining module, configured to determine a candidate flame region in a current frame image; and a processing module, configured to perform the candidate flame region based on a grid algorithm Performing real-time optical flow calculation to obtain an optical flow vector of a pixel of the candidate flame region; and an identifying module configured to determine whether a flame exists according to a distribution characteristic of the optical flow vector of the pixel of the candidate flame region in each direction.

Based on the above technical solution, the method and apparatus for detecting a flame according to an embodiment of the present invention calculates an optical flow vector of a pixel of a candidate flame region by real-time optical flow calculation based on a grid algorithm, and distributes the optical flow vector in each direction according to the optical flow vector. The characteristics determine the presence or absence of a flame and enable accurate and stable detection of the flame in a variety of complex or simple environments.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without paying any creative work.

1 is a schematic flow chart of a method of detecting a flame according to an embodiment of the present invention.

2 is a schematic diagram of vector partitioning in accordance with an embodiment of the present invention.

3 is a schematic illustration of a candidate flame region in accordance with an embodiment of the present invention.

4 is a schematic flow chart of a method for optical flow calculation according to an embodiment of the present invention.

Figure 5 is a schematic block diagram of an apparatus for detecting a flame in accordance with an embodiment of the present invention.

6 is a schematic block diagram of an identification module in accordance with an embodiment of the present invention.

Figure 7 is a schematic illustration of a scene of 20 flame detections in accordance with an embodiment of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are a part of the embodiments of the present invention, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of the present invention. FIG. 1 shows a schematic flow diagram of a method 100 of detecting a flame in accordance with an embodiment of the present invention. As shown in FIG. 1, the method 100 includes:

5110, determining a candidate flame region in the current frame image;

S120, performing real-time optical flow calculation on the candidate flame region based on a grid algorithm, and acquiring an optical flow vector of the pixel of the candidate flame region;

S130. Determine whether a flame exists according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction.

In an embodiment of the invention, the means for detecting the flame performs flame detection based on real-time optical flow calculations. The device for detecting the flame first determines the candidate flame region in the current frame image, and then performs real-time optical flow calculation based on the grid algorithm, obtains the optical flow vector of the pixel of the candidate flame region, and then distributes the characteristics according to the optical flow vector in each direction. To determine if there is a flame. The optical flow calculation based on the grid algorithm can obtain the optical flow vector of each pixel in real time, and the instability and expansion expansion of the flame determine that the distribution of the optical flow vector in the flame region is different from other objects. The characteristics of the optical flow vector are such that the flame can be accurately detected by statistically analyzing the direction distribution characteristics of the optical flow vector.

Therefore, in the method for detecting a flame according to an embodiment of the present invention, an optical flow vector of a pixel of a candidate flame region is obtained by real-time optical flow calculation based on a grid algorithm, and whether a flame exists according to a distribution characteristic of the optical flow vector in each direction is determined. It is able to accurately and stably detect flames in a variety of complex or simple environments.

In S110, the means for detecting the flame determines a candidate flame region in the current frame image.

Since the special color of the flame is an important feature that distinguishes it from other objects, the embodiment of the present invention obtains a candidate flame region by performing color analysis on the image. Optionally, S110 includes:

5111, performing red, green, and blue RGB three-channel color separation on the current frame image, and acquiring pixel values of three channels of each pixel of the current frame image;

5112, determining a pixel point that satisfies the following condition as a pixel point having a flame color characteristic,

R > R _t ,

R > G > B,

S > (255 - R) ^ S _t / R _t ,

Where R is the red dot pixel value of the pixel, G is the green channel pixel value of the pixel, B is the blue channel pixel value of the pixel, Rt is the red channel threshold, S is the saturation of the pixel, and St is the saturation. value; S113. Determine an area including all pixel points having a flame color feature as the candidate flame area.

Specifically, in the embodiment of the present invention, the device for detecting the flame performs RGB three-channel color separation on the current frame image of the monitoring video. If the pixel value of the RGB three-channel of a certain pixel in the current frame image satisfies the above condition, then determining The pixel is a pixel having a flame color characteristic, and then an area including all the pixels having the flame color feature is determined as the candidate flame region. For example, the smallest rectangular area containing all of the pixels having the flame color feature is determined as the candidate flame area for the next optical flow calculation.

At S120, the device for detecting the flame performs real-time optical flow calculation on the candidate flame region based on a grid algorithm, and acquires an optical flow vector of the pixel of the candidate flame region.

The optical flow vector ( ^3,4 , u and V respectively represent the horizontal and vertical velocity components of the pixel point moving from the previous frame image to the next frame image. In the embodiment of the present invention, optionally, according to the current frame The image data of the candidate flame region of the image and the region at the same position in the previous frame image acquires the optical flow vector of the candidate flame region of the current frame image. The following describes in detail how to acquire the optical flow vector.

First, an objective function is established based on two assumptions. When the value of the optical flow vector reflects the real situation, the objective function should theoretically take the minimum value. Then, the optical flow vector can be transformed into a minimized objective function.

The gray value consistency assumption is that the pixel gray value does not change much during the dt time. Use the mathematical formula to describe as:

I(x, y, t) = I(x, y, t + dt) ( 1 ) The gradient consistency assumption, that is, the gradient of each point should be changed smoothly during dt time, which is described by mathematical formula as:

VI(x,y,t) = VI(x,y,t + dt) ( 2 ) Construct the objective function according to the above two consistency assumptions. For example, the objective function can be:

E( ,v) = f (w ^T J (V )w + (\ V \ ² + \ Vv \ ² ))dxdy ( 3 ) is the smoothing operator with p as the standard deviation, J is the structural tensor, V ₃ / denotes the spatiotemporal gradient vector ( y Y ^ w( _y) = (M( _y), v( _y), l) ⁷ is a vector field describing the displacement, "is a tunable coefficient.

Find the minimum value for equation (3), and the corresponding C equation for the optical flow vector to take the minimum value of the objective function (3) using Lagrangian law:

(5) For a system of nonlinear equations, to solve the system of equations, convert it to a system of linear equations: ∑ (J _Ui u _i +J _l2i v _i +J _ni ) = 0

a

(6)

For fast convergence, the embodiment of the present invention uses a Gauss-Seidel iterative algorithm based on a grid algorithm to solve an unknown quantity, the kth number of times

The optical flow vector calculation based on the grid algorithm treats each point on the 2D image as a point on the grid. During the conversion from the fine grid to the coarse grid, the grid points are reduced by one and a half, and the grid size is increased. 1 times. The Gauss-Seidel iterative algorithm is used to solve the unknown quantity on the coarse grid, that is, after obtaining the exact solution, the solution is converted to the fine mesh. In the iterative process of the grid algorithm, the limit operator uses the 4-point averaging method, and the continuation operator uses the interpolation method. The process of converting a coarse mesh to a fine mesh is a process in which the variable becomes large, using an interpolation method (ie, an continuation operator), for example, a vector of 1 point on the coarse mesh is known, and the corresponding fine mesh is placed on it. The four points are assigned to the point vector. The process of transforming a fine mesh to a coarse mesh is a process in which the variable becomes less, using a 4-point averaging method (ie, a limit operator), for example, a vector of 4 points on a fine mesh is known, and the four points are averaged. The value is assigned to a point on the corresponding coarse grid.

Specifically, the V-type multigrid and the nonlinear multigrid algorithm are combined to obtain the fastest convergence speed without increasing the amount of calculation. Use h for the grid size and 2h for the coarse grid size. The algorithm steps are as follows:

1. Initialize the data.

2. If the current mesh is the coarsest mesh, use the Gauss-Seidel algorithm to obtain the solution of the unknown vector of the layer mesh;

Otherwise, execute:

1) On the h grid, iterate using the Gauss-Seidel algorithm to obtain the initial solution;

2) Calculate the margin on the h-layer grid, ue, ve;

3) Execute the limit operator to obtain the margin limit;

4) Set the 2h layer mesh unknown to 0;

5) Execute the Gauss-Seidel iterative algorithm to obtain the solution on the grid;

6) Execute the continuation operator to obtain the solution of the h mesh layer;

7) update the solution;

8) Perform a smooth iteration to reduce the error introduced by the continuation operator;

9) Calculate the margin of the h layer;

10) Execute the restriction operator again to obtain the limit margin;

11) Set the 2h layer mesh unknown to 0;

12) Get the residual by solving the 2h layer mesh layer solution;

13) executing an continuation operator to obtain a solution of the h mesh layer;

14) update the solution;

15) Perform a smooth iteration to reduce the error introduced by the continuation operator.

3. If the number of iterations is complete, exit; otherwise, return to step 2.

In the embodiment of the present invention, the device for detecting the flame uses each pixel point of the candidate flame region of the current frame image as a point on the grid, and obtains the candidate flame region by the iteration of the coarse mesh and the fine mesh. The optical flow vector for each pixel. The Gauss-Seidel iterative algorithm based on the grid algorithm of the embodiment of the invention can quickly converge, so that the optical flow vector can be acquired in real time to detect the flame. Therefore, the method for detecting the flame of the embodiment of the invention has high real-time performance.

In S130, the means for detecting the flame determines whether or not there is a flame based on a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction.

The instability of the flame and the expansion expansion determine the distribution of the optical flow vector in the flame region in all directions with characteristics different from those of other objects. Therefore, the means for detecting the flame can determine whether or not there is a flame based on the distribution characteristics of the optical flow vectors of the pixel points of the candidate flame region in various directions. Optionally, S130 includes:

S131. Determine, according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction, whether the candidate flame region is a real flame region;

S132. Determine whether a flame exists according to whether the candidate flame region is a real flame region. The device for detecting the flame first determines whether the candidate region is a real flame region according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region of the current frame image, that is, whether the real flame exists in the current frame image. The area is then determined whether a flame is present based on whether a real flame region exists in the current frame image.

In S131, the means for detecting the flame determines whether the candidate flame region is a real flame region based on a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction.

As shown in Fig. 2, the direction distribution of the optical flow vector can be divided into four regions, that is, the optical flow vector can be divided into four categories, for example, it can be divided as follows:

1) Right, -— < anole <—

4 4

, ,

2) Tens of thousands,

3) Left, ― < anole≤ π ^ - π < anole≤ -—

4 4

3) 71 j π

4) tens of thousands, < angle < -

4 4

Where angle represents the direction of the optical flow vector.

Considering the characteristics of the flame, the optical flow vector in the flame region will be the most upward, followed by the left and right. Therefore, optionally, the S131 includes: If the optical flow vector of ≤ angle < is greater than the first predetermined threshold in the optical flow vector of the pixel of the candidate flame region, determining that the candidate flame region is a real flame region; or, if the candidate is in the candidate In the optical flow vector of the pixel of the flame region, the optical flow vector of ≤ angle < occupies a specific gravity greater than the first predetermined threshold, and

- - < angle < - < angle < r or - π < angle <―" - The ratio of the optical flow vector to the second predetermined threshold determines that the candidate flame region is the true flame region.

Specifically, after acquiring the optical flow vector of the pixel of the candidate flame region, the device for detecting the flame counts the distribution characteristics of each optical flow vector in each direction, and determines the candidate according to whether the distribution characteristic meets the characteristics of the flame. Whether the flame area is a real flame area. Alternatively, if the specific gravity of the upward optical flow vector exceeds the first predetermined threshold, the candidate flame region may be determined to be a real flame region. The first predetermined threshold can be obtained from a large number of sample statistics in different environments. Optionally, the true flame region may be further determined by combining the optical flow vector with the specific gravity to the left or the right, for example, if the proportion of the upward optical flow vector exceeds the first predetermined threshold, and the light to the left and right If the specific gravity of the flow vector exceeds a second predetermined threshold, then the candidate flame region can be determined to be a true flame region. Similarly, the second predetermined threshold can also be obtained from a large number of sample statistics in different environments.

It should be understood that the manner of determining the true flame region according to the distribution characteristics of the optical flow vector in various directions may have other transformation modes, for example, the ratio of the optical flow vector according to ^ ≤ «^£? ≤ ^

3 3

It is to be determined that all other transformations are intended to be covered by the scope of the present invention.

It should also be understood that when counting optical flow vectors, it may not be necessary to count all optical flow vectors. For example, if the size of an optical flow vector is less than a predetermined value, it is not necessary to count the optical flow vector, or to calculate the optical flow vector at some locations. These statistical methods are also intended to be within the scope of the present invention.

In S132, the means for detecting the flame determines whether or not there is a flame based on whether or not the candidate flame region is a real flame region.

The candidate flame region of the current frame image is the real flame region, that is, the current frame image has a real flame region. The means for detecting the flame may determine whether or not there is a flame based on whether the current frame image has a real flame region, or determine whether or not there is a flame according to the number of frames of the image of the real flame region in the predetermined frame number image.

Therefore, optionally, S132 includes: If the candidate flame zone is a true flame zone, then it is determined that a flame is present.

That is, it is determined whether or not there is a flame based only on whether or not the current frame image has a real flame region.

Optionally, S132 includes:

According to the number of frames of the image of the real flame region in the image within the predetermined number of frames L, it is determined whether or not there is a flame.

In order to improve the accuracy of detecting the flame, the device for detecting the flame determines not only the flame according to whether the current frame image has a real flame region, but also the image of the image of the real flame region in the image within the predetermined number of frames L before the current frame image. Number 1, to determine if there is a flame. Alternatively, if 1/L is greater than the third predetermined threshold, it is determined that there is a flame. That is, if the specific gravity of the number of frames of the real flame in the predetermined number of frames exceeds the third predetermined threshold, it is determined that there is a flame. Similarly, the third predetermined threshold can also be obtained from a large number of sample statistics in different environments. Preferably, the embodiment of the present invention determines the manner of the flame according to the first predetermined threshold, the second predetermined threshold, and the third predetermined threshold, that is, within the predetermined number of frames L, the statistics satisfy the first predetermined threshold and the second predetermined threshold. The ratio of the number of frames to the total number of frames L, if the ratio is higher than the third predetermined threshold, determines that there is a real flame within the number L of frames, and the flame detection is completed.

Therefore, in the method for detecting a flame according to the embodiment of the present invention, whether the flame is present according to the distribution characteristics of the optical flow vectors of the pixel points of the candidate flame region in various directions can improve the accuracy of detecting the flame, and is complicated or simple. The flame can be detected accurately and steadily in the environment.

The method of detecting a flame of an embodiment of the present invention will now be described in more detail by taking a flame sample as an example. It is to be understood that the examples are only intended to assist those skilled in the art to understand some of the possible embodiments of the present invention, and are not to be construed as limiting the scope of the invention.

1. Determine the candidate flame area.

The device for detecting the flame performs color separation on the image obtained from the source video file, and determines a pixel point satisfying the following conditions as a pixel having a flame color characteristic (take 170, take 50),

i? > 170,

R > G > B,

S > (255- ?) * 50 / 170.

The source image is binarized and reconstructed by the above conditions, and the pixel that satisfies the condition is set to white, and vice versa. The minimum rectangular area containing all the white pixels is then framed, and the source image area corresponding to the rectangular area is determined as the candidate flame area (as shown in FIG. 3). 2. Real-time optical flow calculation based on grid algorithm.

As shown in FIG. 4, the device for detecting the flame takes as input the image data of the determined candidate flame region and the region at the same position in the previous frame image, and performs real-time optical flow calculation based on the grid algorithm, each of the candidate flame regions. The pixel is used as a point on the grid, and the optical flow vector of each pixel of the candidate flame region is obtained by transforming the layer coarse grid and the fine mesh to obtain a light flow vector diagram.

The optical flow vector map reflects the optical flow vector of each pixel, and stores the vector information with the image data of each pixel in the grayscale image. In order to visually display the optical flow vector of the candidate flame region, a vector diagram is drawn in units of a grid of 10*10 pixels, and each arrow in the vector diagram below FIG. 4 is a vector of the 10*10 pixel grid.

3. Statistical optical flow vector characteristics.

The device for detecting the flame counts the distribution of the optical flow vectors of the respective pixels in the four directions as shown in FIG. In the vector diagram shown in Figure 4, the vector with the vector direction above accounts for nearly 60% of the global weight, and the vector with the vector direction to the left or the right accounts for nearly 30% of the global weight, which is consistent with the characteristics of the flame. Can be determined as a real flame area. Finally, the number of image frames existing in the real flame region according to the predetermined number of frames satisfies a predetermined threshold, and it is determined that there is a flame, and the flame detection is completed.

Therefore, in the method for detecting a flame according to an embodiment of the present invention, an optical flow vector of a pixel of a candidate flame region is obtained by real-time optical flow calculation based on a grid algorithm, and whether a flame exists according to a distribution characteristic of the optical flow vector in each direction is determined. , able to accurately and stably detect the flame.

A method of detecting a flame according to an embodiment of the present invention is described in detail above with reference to FIGS. 1 through 4, and an apparatus for detecting a flame according to an embodiment of the present invention will be described below with reference to FIGS. 5 and 6.

Figure 5 shows a schematic block diagram of an apparatus 500 for detecting a flame in accordance with an embodiment of the present invention. As shown in Figure 5, the apparatus 500 includes:

a determining module 510, configured to determine a candidate flame region in the current frame image;

The processing module 520 is configured to perform real-time optical flow calculation on the candidate flame region based on a grid algorithm, and acquire an optical flow vector of the pixel of the candidate flame region;

The identifying module 530 is configured to determine whether a flame exists according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction.

The device for detecting a flame according to an embodiment of the present invention calculates an optical flow vector of a pixel of a candidate flame region by real-time optical flow calculation based on a grid algorithm, and distributes the light flow vector in each direction according to the optical flow vector. Determine if a flame is present and be able to accurately and stably detect the flame in a variety of complex or simple environments. In the embodiment of the present invention, the determining module 510 includes:

a color separation unit, configured to perform red, green, and blue RGB three-channel color separation on the current frame image, and obtain pixel values of three channels of each pixel of the current frame image;

a feature pixel determining unit configured to determine a pixel point that satisfies the following condition as a pixel point having a flame color characteristic,

R > R _t ,

R > G > B,

S > (255 - R) ^ S _t / R

Where R is the red dot pixel value of the pixel, G is the green channel pixel value of the pixel, B is the blue channel pixel value of the pixel, Rt is the red channel threshold, S is the saturation of the pixel, and St is the saturation. Value

A candidate flame region determining unit is configured to determine an area including all pixel points having a flame color characteristic as the candidate flame region.

In the embodiment of the present invention, the processing module 520 is specifically configured to: use each pixel point of the candidate flame region as a point on the mesh, and obtain the transformation iteration of the coarse mesh and the fine mesh to obtain The optical flow vector for each pixel of the candidate flame region.

The Gauss-Seidel iterative algorithm based on the grid algorithm of the embodiment of the present invention can quickly converge, so that the optical flow vector can be acquired in real time to detect the flame. Therefore, the device for detecting the flame of the embodiment of the invention has high real-time performance.

In the embodiment of the present invention, as shown in FIG. 6, the identification module 530 includes: a determining unit 531, configured to determine, according to a distribution characteristic of optical flow vectors of pixel points of the candidate flame region in various directions, Whether the candidate flame region is a real flame region;

The identifying unit 532 is configured to determine whether a flame exists according to whether the candidate flame region is a real flame region.

Optionally, the determining unit 531 includes:

a first determining subunit, configured to be in an optical flow vector of a pixel point of the candidate flame region,

If the proportion of the optical flow vector of ^ ≤ angle ≤ ^ is greater than the first predetermined threshold, then the candidate flame region is determined to be the real flame region, and angle represents the direction of the optical flow vector. Optionally, the determining unit 531 includes:

a second determining subunit, configured to: in the optical flow vector of the pixel of the candidate flame region, the optical flow vector of ^ < angle occupies a specific gravity greater than a first predetermined threshold, and

- - < angle < - < angle < r or - π < angle <―" - the proportion of the optical flow vector is greater than the second predetermined threshold, then the candidate flame region is determined to be the real flame region, and angle represents the optical flow vector The direction.

In the embodiment of the present invention, optionally, the identifying unit 532 includes:

The first identifying subunit is configured to determine whether there is a flame according to the number of frames 1 of the image in which the real flame region exists in the image within the predetermined number of frames L.

Optionally, the first identifying subunit is specifically configured to determine that a flame exists if 1/L is greater than a third predetermined threshold.

Optionally, the identifying unit 532 includes:

The second identification subunit is configured to determine that there is a flame if the candidate flame region is a real flame region.

The apparatus 500 for detecting a flame according to an embodiment of the present invention may correspond to an execution body of the method of detecting a flame in the embodiment of the present invention, and the above-described and other operations and/or functions of the respective modules in the apparatus 500 are respectively for implementing FIG. 1 to FIG. The corresponding flow of each method in 4, for brevity, will not be repeated here.

The device for detecting a flame according to an embodiment of the present invention can improve the accuracy of detecting a flame by using a distribution characteristic of optical flow vectors of pixel points of candidate flame regions in various directions to improve the accuracy of detecting flames in various complicated or simple environments. The flame can be detected accurately and stably.

Table 1 shows the results of flame detection of 20 scenes in the embodiment of the present invention and the flame detection technique in the prior art by establishing a hidden Markov model to characterize the flame flicker. Figure 7 is a schematic diagram of the 20 scenes. The 20 scene numbers are 1-20, where 1-14 contains the flame scene, as the positive sample, and 15-20 is the non-flame scene. As the negative sample, the monitoring equipment of No. 1-4 is very stable without any displacement. Monitoring equipment No. 5-14 has slight displacement or large displacement. Table 1

It can be seen from Table 1: In the case that the monitoring environment is relatively stable, the prior art scheme also has High detection rate. However, in the case where the monitoring environment is complicated or the monitoring device is slightly shaken, the detection accuracy of the prior art scheme is greatly reduced and there is a certain degree of false alarm (for example, in the scene 15, the monitoring equipment of the underground parking lot is slightly shaken, resulting in warning The light from the lamp is falsely reported as a flame). However, for the technical solution of the embodiment of the present invention, due to the particularity of the flame optical flow property, the detection accuracy and the extremely low false alarm rate are high in a simple or complicated scene.

Therefore, the method and apparatus for detecting a flame according to an embodiment of the present invention calculates an optical flow vector of a pixel point of a candidate flame region by real-time optical flow calculation based on a grid algorithm, and determines whether the optical flow vector has a distribution characteristic in each direction according to the optical flow vector. There is a flame that can accurately and stably detect the flame in a variety of complex or simple environments.

It should be understood that in the embodiment of the present invention, the term "and/or" is merely an association describing the associated object, indicating that there may be three relationships. For example, A and / or B, can mean: A exists separately, there are A and B, and there are three cases of B alone. In addition, the character "/" in this article generally means that the contextual object is an "or" relationship.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both, for clarity of hardware and software. Interchangeability, the composition and steps of the various examples have been generally described in terms of function in the above description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.

A person skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding processes in the foregoing method embodiments, and details are not described herein again.

In the several embodiments provided herein, it should be understood that the disclosed systems, devices, and methods may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, also It can be electrical, mechanical or other form of connection. The components displayed by the unit may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the embodiments of the present invention.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention contributes in essence or to the prior art, or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium. A number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and the like, which can store program codes. .

The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any equivalent person can be easily conceived within the technical scope of the present invention. Modifications or substitutions are intended to be included within the scope of the invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

Rights request

A method for detecting a flame, comprising:

Determining candidate flame regions in the current frame image;

Performing real-time optical flow calculation on the candidate flame region based on a grid algorithm, and acquiring an optical flow vector of a pixel of the candidate flame region;

The presence or absence of a flame is determined based on the distribution characteristics of the optical flow vectors of the pixel points of the candidate flame regions in respective directions.

The method according to claim 1, wherein the determining, according to a distribution characteristic of the optical flow vector of the pixel points of the candidate flame region in each direction, whether a flame is present, comprises: according to the candidate flame a distribution characteristic of the optical flow vector of the pixel of the region in each direction, determining whether the candidate flame region is a real flame region;

Whether or not there is a flame is determined according to whether the candidate flame region is a real flame region.

The method according to claim 2, wherein the determining, according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction, whether the candidate flame region is a real flame region, Includes:

If in the optical flow vector of the pixel of the candidate flame region, the optical flow vector of ≤ angle occupies a specific gravity greater than the first predetermined threshold, determining that the candidate flame region is a real flame region, and angle represents an optical flow vector Orientation; or,

If the optical flow vector of the pixel of the candidate flame region occupies a specific gravity greater than the first predetermined threshold,

- - < angle < - < angle ≤ r or - π < angle ≤ "" - the proportion of the optical flow vector is greater than the second predetermined threshold, then the candidate flame region is determined to be the real flame region, and angle represents the optical flow Vector direction.

The method according to claim 2 or 3, wherein the determining whether the flame is present according to whether the candidate flame region is a real flame region comprises:

According to the number of frames of the image of the real flame region in the image within the predetermined number of frames, it is determined whether or not there is a flame.

The method according to claim 4, wherein the determining, according to the number of frames of the image of the real flame region, in the image within the predetermined number of frames, determining whether a flame is present comprises: If the IIL is greater than the third predetermined threshold, then it is determined that there is a flame.

If the candidate flame zone is a true flame zone, then it is determined that a flame is present.

The method according to claim 1 or 2, wherein the determining the candidate flame region in the current frame image comprises:

Performing red, green, and blue RGB three-channel color separation on the current frame image, and acquiring pixel values of three channels of each pixel of the current frame image;

A pixel point that satisfies the following condition is determined as a pixel point having a flame color characteristic,

R > R _t ,

R > G > B,

S > (255 - R) ^ S _t / R

Wherein, represents a red dot pixel value of a pixel point, G represents a green channel pixel value of a pixel point, B represents a pixel value of a pixel blue channel, represents a red channel threshold, S represents a saturation of a pixel, and represents a saturation threshold;

An area including all pixel points having a flame color characteristic is determined as the candidate flame area.

The method according to claim 1 or 2, wherein the grid-based algorithm performs real-time optical flow calculation on the candidate flame region, and acquires an optical flow vector of a pixel of the candidate flame region, including :

Each pixel of the candidate flame region is taken as a point on the grid;

The optical flow vector of each pixel of the candidate flame region is obtained by a transformation iteration of the coarse mesh and the fine mesh.

9. A device for detecting a flame, comprising:

a determining module, configured to determine a candidate flame region in the current frame image;

a processing module, configured to perform real-time optical flow calculation on the candidate flame region based on a grid algorithm, and obtain an optical flow vector of a pixel of the candidate flame region;

And an identification module, configured to determine whether a flame exists according to a distribution characteristic of the optical flow vector of the pixel point of the candidate flame region in each direction.

The device according to claim 9, wherein the identification module comprises: a determining unit, configured to use optical flow vectors of pixel points of the candidate flame region in various directions a distribution characteristic, determining whether the candidate flame region is a real flame region;

And an identifying unit, configured to determine whether a flame exists according to whether the candidate flame region is a real flame region.

The device according to claim 10, wherein the determining unit comprises: a first determining subunit, configured to: ≤ g/e ≤ in an optical flow vector of a pixel point of the candidate flame region If the proportion of the optical flow vector is greater than the first predetermined threshold, determining that the candidate flame region is a real flame region, indicating a direction of the optical flow vector; or

a second determining subunit, configured to be in an optical flow vector of a pixel point of the candidate flame region,

^ < angle of the optical flow vector occupies a larger proportion than the first predetermined threshold, and - - < angle < ― < angle < τ or - π < angle < ― " ^ " the proportion of the optical flow vector is greater than the Two predetermined thresholds determine that the candidate flame region is a true flame region and angle represents a direction of the optical flow vector.

The apparatus according to claim 10 or 11, wherein the identification unit comprises: a first identification sub-unit, configured to: according to the number of frames of the image of the real flame region in the image within the predetermined number of frames/ , to determine if there is a flame.

13. Apparatus according to claim 12 wherein said first identifying subunit is operative to determine that a flame is present if // is greater than a third predetermined threshold.

14. Apparatus according to claim 10 or claim 11 wherein the identification unit comprises: a second identification subunit for determining the presence of a flame if the candidate flame zone is a true flame zone.

The device according to claim 9 or 10, wherein the determining module comprises: a color separating unit, configured to perform red, green, blue, and RGB three-channel color separation on the current frame image to obtain the current frame. The pixel value of three channels per pixel of the image;

R > R _t ,

R > G > B,

S > (255 - R) ^ S _t / R

Wherein, the pixel value of the red channel of the pixel is represented, G represents the pixel value of the green channel of the pixel, B represents the pixel value of the blue channel of the pixel, represents the red channel threshold, and S represents the saturation of the pixel. & indicates the saturation threshold;

A candidate flame area determining unit is configured to determine an area including all pixel points having a flame color characteristic as the candidate flame area.

The device according to claim 9 or 10, wherein the processing module is configured to use each pixel of the candidate flame region as a point on the grid, through a coarse mesh and a thin The transformation iteration of the mesh acquires an optical flow vector for each pixel of the candidate flame region.