Summary of the invention
The object of the present invention is to provide a kind of methods of continuous monitoring boiler combustion situation, have precision height, reaction time
Short advantage.
To achieve the above object, the present invention provides following schemes:
The invention discloses a kind of methods of continuous monitoring boiler combustion situation, which comprises
Acquire the vision signal of the furnace flame in combustion zone to be monitored in the full load situation lower predetermined time;
The time-series image of flame kernel is obtained according to the vision signal;
It is calculated according to the time interval of frame image every in the time-series image, the focal length of every frame image and size of burner hearth
The pixel number of every frame image out, the pixel number of more all time-series images determine image scale referring to image;
Hearth combustion image to be identified is obtained in real time, according to the hearth combustion image to be identified and described image ratio
Example ruler determines the monitoring region of the hearth combustion image to be identified referring to image;
According to the characteristics of image of the hearth combustion image to be identified, using supporting vector machine model to described to be identified
Hearth combustion image classify;
According to the characteristics of image and type of the hearth combustion image to be identified, it is described for using supporting vector machine model
Hearth combustion image to be identified determines flame identification algorithm, in the monitoring region for calculating the hearth combustion image to be identified
The two-dimensional coordinate of flame central position;
The two-dimensional coordinate is converted into the corresponding flame kernel of the hearth combustion image to be identified in burner hearth
Position.
Optionally, the time-series image for obtaining flame kernel according to the vision signal specifically includes: using
Adobe Premiere software carries out the time-series image that processing obtains flame kernel to the vision signal.
Optionally, the hearth combustion image to be identified is pressed into HSV (Hue, Saturation, Value) color space
3 channels are divided into, the average brightness value and mean square deviation in each channel of hearth combustion image to be identified are calculated separately, it will
The average brightness value and mean square deviation in each channel of hearth combustion image to be identified are as the hearth combustion to be identified
The described image feature of image.
Optionally, the classification of the hearth combustion image to be identified includes the image and Abnormal combustion of normal combustion
Image, the image of the normal combustion are that combustion state fires the image of thermic load state interval to minimum steady at full capacity in burner hearth,
The image of the Abnormal combustion is more than combustion state at full capacity or lower than the image of minimum steady combustion thermic load state.
Optionally, described that institute is determined referring to image according to the hearth combustion image to be identified and described image scale bar
The monitoring region for stating hearth combustion image to be identified, specifically includes:
Step 1041: judging whether the hearth combustion to be identified is the image of normal combustion, if going to step
1042, step 1043 is gone to if not;
Step 1042: using described image scale bar referring to image as the first picture, by the hearth combustion to be identified
Image is as second picture;First picture and the second picture are subjected to gamma correction respectively;By the second picture
The picture in the channel brightness V after the segmentation of hsv color space is as third picture;According to first picture after gamma correction
Part is carried out certainly to the third picture with the absolute value of the difference of the picture pixels matrix of the second picture after gamma correction
Thresholding processing, closing operation of mathematical morphology and dilation operation are adapted to, and the burner hearth combustion to be identified is obtained by profile operation
Burn the monitoring region of image;
Step 1043: the average brightness of the hearth combustion image irradiation to be identified is calculated, if average brightness is greater than 90,
Then go to the step 1042;Otherwise, hsv color segmentation is carried out to the hearth combustion image to be identified, takes the channel brightness V
Picture carry out intermediate value fuzzy, the processing of histogram equalization processing, local auto-adaptive thresholding, closing operation of mathematical morphology and expansion
Operation, and obtain by profile operation the monitoring region of the hearth combustion image to be identified.
Optionally, the pixel number of more all time-series images determine image scale referring to image it
Before, it specifically includes: binary conversion treatment being carried out to frame image every in the time-series image, every frame after calculating binary conversion treatment
The pixel number of image.
Optionally, the vision signal of the furnace flame in the acquisition combustion zone to be monitored under full load situation is specifically wrapped
It includes: obtaining the vision signal of furnace flame using furnace flame monitoring system.
Optionally, which is characterized in that the focal length is the focal length of periscope pipe in the furnace flame monitoring system.
The summary of the invention provided according to the present invention, the invention discloses following technical effects: video of the present invention to acquisition
Signal carries out software processing, and determines that scale bar referring to image, utilizes according to software treated image data and size of burner hearth
Supporting vector machine model carries out classification and the selection of flame kernel recognizer, present invention benefit to the images to be recognized obtained in real time
Classification and the selection of flame kernel computational algorithm are carried out to image with supporting vector machine model, it can be compared with using the stability of algorithm
Inverted image, refraction and the dirty influence to hearth combustion situation of burner hearth ash are solved well, to improve in identification burner hearth flame in real time
The accuracy of the position of the heart.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of methods of continuous monitoring boiler combustion situation, realize burner hearth flame center
The real-time monitoring of position has the advantages that precision is high, the reaction time is short.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Fig. 1 is a kind of method flow schematic diagram of continuous monitoring boiler combustion situation of the embodiment of the present invention, as shown in Figure 1,
A kind of continuous method of monitoring boiler combustion situation provided by the invention includes:
Step 101: acquiring the video letter of the furnace flame in combustion zone to be monitored in the full load situation lower predetermined time
Number;
Step 102: the time-series image of flame kernel is obtained according to the vision signal;
Step 103: according to the time interval of frame image every in the time-series image, the focal length and burner hearth of every frame image
Size calculates the pixel number of every frame image, the pixel number of more all time-series images, by the figure that flame is most full
As being determined as image scale referring to image;Crucial scale to image scale referring to image is demarcated.
Step 104: hearth combustion image to be identified is obtained in real time, according to the hearth combustion image to be identified and institute
It states image scale and determines the monitoring region of the hearth combustion image to be identified referring to image, and calculate monitoring region phase
The X-axis of corresponding two-dimensional coordinate and the full scale value of Y-axis;The size and described image scale bar in the monitoring region are referring to image
Size it is consistent;
Step 105: according to the characteristics of image of the hearth combustion image to be identified, using supporting vector machine model to institute
Hearth combustion image to be identified is stated to classify;
Step 106: according to the characteristics of image and type of the hearth combustion image to be identified, using support vector machines mould
Type is that the hearth combustion image to be identified determines flame identification algorithm, calculates the prison of the hearth combustion image to be identified
Survey the two-dimensional coordinate of region Flame center;Flame kernel refers to the most bright point of image Flame;
Fig. 3 is that flame location coordinate scale of the present invention demarcates schematic diagram, as shown in Figure 3: remembering hearth combustion figure to be identified
The full scale coordinate value in the monitoring region of picture is (M, N);Current hearth combustion image flame coordinate to be identified is (m, n);X is sat
Coordinate value is that the coordinate points of 0, M/3,2M/3 and 3M/3 are expressed as M1, M2, M3 and M4, coordinate value in Y-coordinate axle on parameter
Be 0, N/3, the point of 2N/3,3N/3 are expressed as point N1, N2, N3 and N4, and are demarcated by image scale, in X-axis from
Three sections are divided into left-to-right, Y-axis from bottom to up:
If coordinate value m is located at first segment in X-axis in current hearth combustion image X-axis to be identified, flame kernel is in X-axis
On coordinate value: U1=(M2-m) * (M/3)/(M2-M1)=M2-m;
If coordinate value m is located at second segment in X-axis in current hearth combustion image X-axis to be identified, flame kernel is in X-axis
On coordinate value: U2=M/3+ (M3-m) * (M/3)/(M3-M2)=M/3+ (M3-m);
If coordinate value m third section in X-axis in current hearth combustion image X-axis to be identified, flame kernel is in X-axis
Coordinate value: U3=2M/3+ (M4-m) * (M/3)/(M4-M3)=2M/3+ (M4-m);
If coordinate value n is located at first segment in Y-axis in current hearth combustion image Y-axis to be identified, flame kernel is in Y-axis
On coordinate value: V1=(N2-n) * (N/3)/(N2-N1)=N2-n;
If coordinate value n is located at second segment in Y-axis in current hearth combustion image Y-axis to be identified, flame kernel is in Y-axis
On coordinate value: V2=N/3+ (N3-n) * (N/3)/(N3-N2)=N/3+ (N3-n);
If coordinate value n is located at third section in Y-axis in current hearth combustion image Y-axis to be identified, flame kernel is in Y-axis
On coordinate value: V3=2N/3+ (N4-n) * (N/3)/(N4-N3)=2N/3+ (N4-n);
Step 107: the two-dimensional coordinate being converted into the corresponding flame kernel of the hearth combustion image to be identified and is existed
Position in burner hearth.
Wherein, step 101, specifically further include: the burner hearth fire in the acquisition combustion zone to be monitored under full load situation
The vision signal of flame specifically includes: the vision signal of furnace flame is obtained using furnace flame monitoring system.Combustion zone to be monitored
Full load situation refers to that Boiler Steam yield reaches combustion conditions when nameplate provides metered flow in domain, i.e., fire inside burner hearth
When flame maximum intensity.
Wherein, step 102, specifically further include: the vision signal is handled using Adobe Premiere software
Obtain the time-series image of flame kernel.Time-series image refers to a series of furnace flame combustion cases changed over time
Image, by time-series image it can be seen that flame height, flame kernel change with time situation.
Wherein, step 103, specifically further include: the pixel number of more all time-series images determines image
Scale bar carries out binary conversion treatment to frame image every in the time-series image, calculates binary conversion treatment referring to before image
The pixel number of every frame image afterwards:
Proportionate relationship, i.e. image scaled are judged by calculating the corresponding pixel number in image Flame section (longitudal section)
Ruler.Judgment method: server is handled the vision signal through Adobe Premiere software, obtains time-series image,
Carry out image binaryzation:
As f (m, n) >=T, f (m, n)=255;
As f (m, n) < T, f (m, n)=0;
T is selected luminance threshold in formula, is set as 60 herein;F (m, n) is at time-series image processing midpoint (m, n)
Brightness value.
The time interval of known every frame image, periscope pipe focal length, size of burner hearth and periscope pipe in burner hearth institute
The position at place, is subject to above-mentioned parameter, works out corresponding image analysis program to the figure after binary conversion treatment using MATLAB
Calculating as carrying out pixel number.
It wherein, step 104, specifically further include that the classification of the hearth combustion image to be identified includes normal combustion
The image of image and Abnormal combustion, the image of the normal combustion are that combustion state is negative to minimum steady combustion heat at full capacity in burner hearth
The image of lotus state interval, the image of the Abnormal combustion are to fire thermic load more than combustion state at full capacity or lower than minimum steady
The image of state.
Wherein, step 104, specifically further include: be divided into the hearth combustion image to be identified by hsv color space
3 channels, calculate separately the average brightness value and mean square deviation in each channel of hearth combustion image to be identified, will it is described to
The average brightness value and mean square deviation in each channel of hearth combustion image of identification are as the hearth combustion image to be identified
Described image feature.
It is described according to the hearth combustion image to be identified and described image scale bar referring to image determine described in wait know
The monitoring region of other hearth combustion image, specifically includes:
Step 1041: judging whether the hearth combustion to be identified is the image of normal combustion, if going to step
1042, step 1043 is gone to if not;
Step 1042: using described image scale bar referring to image as the first picture, by the hearth combustion to be identified
Image is as second picture;First picture and the second picture are subjected to gamma correction respectively;By the second picture
The picture in the channel brightness V after the segmentation of hsv color space is as third picture;According to first picture after gamma correction
Part is carried out certainly to the third picture with the absolute value of the difference of the picture pixels matrix of the second picture after gamma correction
Thresholding processing, closing operation of mathematical morphology and dilation operation are adapted to, and the burner hearth combustion to be identified is obtained by profile operation
Burn the monitoring region of image;
Gamma correction can detect dark parts and light-colored part in picture signal, and increase the two ratio, thus
Improve image comparison effect.
Step 1043: the average brightness of the hearth combustion image irradiation to be identified is calculated, if average brightness is greater than 90,
Then go to the step 1042;Otherwise, hsv color segmentation is carried out to the hearth combustion image to be identified, takes the channel brightness V
Picture carry out intermediate value fuzzy, the processing of histogram equalization processing, local auto-adaptive thresholding, closing operation of mathematical morphology and expansion
Operation, and obtain by profile operation the monitoring region of the hearth combustion image to be identified.
The focal length is the focal length of periscope pipe in the furnace flame monitoring system.
Fig. 2 is video signal acquisition device structure chart of the present invention, as shown in Fig. 2, electronic actuator 205 is by periscope pipe
203 carry out the combustion position inside record burner hearth inside 201 tapping of furnace wall push-in burner hearth, wherein connecting plate 202 is for fixing
Electronic actuator 205, for protecting periscope pipe 203, power supply box 208 is used to provide power supply for device camera protecting cover 206,
Server 209 is used to receive the vision signal of the acquisition of periscope pipe 203 and carries out the data processing of vision signal, blowing pipe 207
Flying dust inside burner hearth is prevented to be attached on periscope pipe 203, iris group 204 has focusing function, to adapt under different load
Combustion position.
The present invention carries out software processing to the vision signal of acquisition, and according to software treated image data and burner hearth ruler
Very little determining scale bar carries out in classification and flame the images to be recognized obtained in real time referring to image, using supporting vector machine model
The selection of heart recognizer can preferably solve inverted image, refraction and burner hearth ash dirt to hearth combustion using the stability of algorithm
The influence of situation, to improve the accuracy of the position at identification burner hearth flame center in real time;The present invention needs not rely on itself
Accuracy to the great object of reference of recognition result disturbance degree and precognition reference height, furnace arch, furnace nose angle, camera level away from
From and height;Artificial subjective factor can be overcome, the defects of operand is big, have precision is high, reaction in time, manipulation intelligence, prison
The advantages of surveying the operation is stable.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention;At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.