CN112052912A - Intelligent flame combustion state identification method for fire-fighting robot - Google Patents
Intelligent flame combustion state identification method for fire-fighting robot Download PDFInfo
- Publication number
- CN112052912A CN112052912A CN202011010733.5A CN202011010733A CN112052912A CN 112052912 A CN112052912 A CN 112052912A CN 202011010733 A CN202011010733 A CN 202011010733A CN 112052912 A CN112052912 A CN 112052912A
- Authority
- CN
- China
- Prior art keywords
- flame
- firefly
- improved
- image
- fire
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000002485 combustion reaction Methods 0.000 title claims abstract description 29
- 241000254158 Lampyridae Species 0.000 claims abstract description 80
- 230000011218 segmentation Effects 0.000 claims abstract description 34
- 238000012706 support-vector machine Methods 0.000 claims abstract description 34
- 238000012545 processing Methods 0.000 claims abstract description 27
- 238000005457 optimization Methods 0.000 claims abstract description 23
- 238000013519 translation Methods 0.000 claims abstract description 19
- 239000002131 composite material Substances 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims description 18
- 230000014509 gene expression Effects 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 15
- 150000001875 compounds Chemical class 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 4
- 238000003708 edge detection Methods 0.000 claims description 4
- 230000031700 light absorption Effects 0.000 claims description 4
- 230000006978 adaptation Effects 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 2
- 230000000007 visual effect Effects 0.000 abstract description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000010355 oscillation Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an intelligent flame combustion state identification method for a fire-fighting robot, which comprises the following steps: carrying out segmentation pretreatment on an original image; calculating radial Tchebichef moment rotation, translation and scaling invariants of the segmentation processing image and the edge image, and constructing a multi-feature composite vector; inputting the multi-feature composite vector into a pre-established and trained improved firefly algorithm-wavelet support vector machine model, and finally obtaining a classification recognition result of the flame combustion state; an improved step size factor and an improved attraction degree are introduced into a firefly algorithm by an improved firefly algorithm-wavelet support vector machine model, iterative search optimization is carried out on wavelet support vector machine kernel parameters, and optimal kernel parameters are obtained. The method is not influenced by the position and the direction of the flame in a visual field scene and the distance between the flame and the image equipment, has high algorithm optimization efficiency and strong recognition performance, can effectively recognize and classify the combustion state of the flame, and assists the fire-fighting robot to perform subsequent targeted fire-fighting operation.
Description
Technical Field
The invention relates to the field of fire-fighting robots, in particular to an intelligent flame identification method for a fire-fighting robot.
Background
With the rapid development of national economy and urbanization construction, various large-scale chemical product storage houses, high-rise building buildings and large-span commercial buildings are scaled. The buildings generally have the characteristics of large personnel mobility, more stored combustible substances, complicated internal channels and the like, and once a fire disaster happens, the buildings are very easy to cause national property loss and casualties. As an intelligent fire fighting means, the fire-fighting robot can replace a fireman to complete reconnaissance and information feedback of a fire scene environment when facing the unknown complex fire scene environment, and carry out fire extinguishing work at the first time. In the process of fire extinguishing work performed by the fire-fighting robot, accurate flame identification is the most critical.
Flame image recognition is a novel image recognition detection technology, intelligent analysis is carried out on images by adopting an advanced recognition algorithm, characteristics such as flame motion, edge contour, color, space difference and the like are extracted, and the flame is directly recognized in real time. For example, patent CN104504382B finds the highest point and the center of gravity of the flame through an internal and external flame extraction algorithm, records RGB values on coordinates of the two points, compares the RGB values with a standard flame RGB feature library to obtain a matching value, and finally determines whether the image is a flame image according to the size of the matching value. For another example, patent CN101826153A detects the motion profile of the current frame and extracts the motion region, and performs uniform color-time domain-space domain filtering on the pixels in the motion region to mark flame pixels, and determines whether to start fire alarm according to the number of flame pixels. In the patent CN104766094B, when a fire disaster does not occur, the numbers of R values and G values are extracted as the distinguishing features, and when the numbers of R values and G values in the monitored image to be detected are obviously abnormal, the fire condition is judged according to the abnormal condition.
However, as with the above methods, most of the existing methods can only identify whether one image contains flames, and cannot identify key fire information such as the combustion state of the flames, and cannot assist the fire-fighting robot to make a perfect fire-fighting scheme. Moreover, these flame recognition methods are very sensitive to rotation, translation and scaling of the image, are susceptible to the location, orientation and distance between the flame and the image device in the scene of the fire, and are not suitable for use in fire fighting robots that are often in motion.
Disclosure of Invention
The invention aims to provide an intelligent flame combustion state identification method for a fire-fighting robot, which overcomes the defects that the combustion state cannot be identified, the visual angle for processing images is single, and the processing capability of the image for sensitive problems such as image rotation, translation and scaling is lacking in the prior art.
The purpose of the invention can be realized by the following technical scheme, comprising the following steps:
image acquisition: acquiring a series of candidate images by using a camera to construct an image data set, and dividing the acquired image data set into a training sample set and a testing sample set;
segmentation pretreatment: processing original images of the training sample set and the test sample set by adopting a color criterion segmentation rule based on a YCbCr color space to obtain corresponding segmentation processing images, then carrying out edge detection processing on the segmentation processing images by using a Canny operator, and extracting the outline of a suspected flame area to obtain an edge image;
and (3) flame feature extraction: calculating rotation, translation and scaling invariants of radial Tchebichef moments of the segmentation processing image and the edge image, respectively obtaining an area moment invariant and a contour moment invariant, and combining the area moment invariants and the contour moment invariants to construct a multi-feature composite vector for describing flame features;
the improved firefly algorithm-wavelet support vector machine is constructed as follows: initializing a pre-established improved firefly algorithm based on the multi-feature composite vector of the training sample set, and performing iterative search optimization on the kernel parameters of the wavelet support vector machine in the firefly transfer process through the introduced improved step-size factor and the improved attraction degree until the preset maximum iteration times are reached to obtain the optimal kernel parameters, so as to construct an improved firefly algorithm-wavelet support vector machine model;
and (3) classifying and identifying the flame combustion state: and inputting the multi-feature composite vector of the test sample set into a pre-established and trained improved firefly algorithm-wavelet support vector machine model, predicting the classification label of the class to which the test sample belongs, and finally obtaining the classification recognition result of the flame combustion state.
Further, in the segmentation preprocessing process, the expression of the color criterion segmentation rule based on the YCbCr color space is adopted as follows:
in the formula, I (x, y) is a segmentation processing image, if a suspected flame pixel point exists at the original image coordinate (x, y), the gray value at the corresponding coordinate is 1, otherwise, the gray value is 0; y ismeanThe average value of Y channel components of a single original image is obtained; y (x, Y), Cb (x, Y) and Cr (x, Y) are respectively the gray value of the luminance component, the gray value of the blue chrominance component and the gray value of the red chrominance component of the pixel point at the coordinate (x, Y) of the original image; λ is an adaptation coefficient, λ > 1; s is the total number of all pixels in a single original image.
Further, in the flame feature extraction process, an expression of extracting flame features by using rotation, translation and scaling invariant of the radial Tchebichef moment is as follows:
in the formula (I), the compound is shown in the specification,rotation, translation and scaling invariants for the radial Tchebichef moment; cpqA radial Tchebichef moment of order p + q; t is00A radial Tchebichef moment of zero order; n is the maximum resolution used in the angular direction; ρ (p, N) is the orthonormal squared norm;a regularized Tchebichef polynomial of order p and length N; m is N/2, and N is the side length of the image; q is a non-negative integer, and q is not more than N-1; g (r, theta) is the input image after translation and mapping; theta is 2 pi l/n, l is more than or equal to 0 and less than or equal to n-1, and l is an integer; r varies from 0 to N/2 and theta varies from 0 to 2 pi.
Further, the kernel function of the wavelet support vector machine adopts a Morlet wavelet function.
Furthermore, in the iterative search optimization process, the improved step factor definition formula is adopted in the improved firefly algorithm to realize the self-adaptive dynamic adjustment of the search step. In the initial stage of iterative search of the algorithm, in order to increase the search space of the firefly individuals, improve the optimization efficiency of the algorithm, avoid local search near the self position in a random mode and increase the step factor value; in the later iteration stage of the algorithm, because random disturbance terms exist all the time, firefly individuals are difficult to coincide with optimal individuals, and in order to improve the convergence rate of the algorithm and prevent repeated oscillation near the optimal feasible solution, the step factor value is reduced at the moment. The improved step-size factor defines the expression of the formula as follows:
in the formula (I), the compound is shown in the specification,the step size factor is the step size factor of the improved n +1 th iteration; n is the current iteration number;is the final value of the step factor;is the initial value of the step factor; n ismaxIs the maximum number of iterations.
Further, in the iterative search optimization process, the improved attraction degree is adopted to define a formula by the improved firefly algorithm. At the initial stage of the algorithm, the spatial distance r between some firefly individuals is caused by the over-dispersion of population distributionijTo reduce the influence of the spatial distance on the attraction level, the search space for individual fireflies in the initial stage of the algorithm operation is further increased, the global optimization capability and the optimization efficiency are again improved, and the basic attraction level other than zero is introduced as the minimum value of the attraction level. The expression of the improved attraction definition formula is as follows:
in the formula (I), the compound is shown in the specification,the method is to improve the attraction degree of the firefly i to the firefly j; rho0The maximum attraction of the firefly i to other individuals; alpha is light absorption coefficient, alpha belongs to [0.1,10 ]];ρbIs basically attractive and is not influenced by medium absorption in the transmission process.
Further, in the iterative search optimization process of the improved firefly algorithm, the expression of the position update formula of the firefly i is as follows:
in the formula (I), the compound is shown in the specification,the space position coordinate vector of the firefly i in the (n + 1) th iteration is obtained; rand is in [0,1 ]]Random numbers are uniformly distributed in the interval.
Compared with the prior art, the invention has the following advantages:
(1) the flame characteristics in the image are extracted by adopting the invariants of rotation, translation and scaling of the radial Tchebichef moment, so that the flame image can be accurately and stably identified, and the flame image is insensitive to the rotation, translation and scaling of the image; and a multi-feature composite vector containing area invariants and contour moment invariants is constructed, so that the global and local features of the flame can be fully utilized.
(2) The step factor and the attractiveness of the firefly algorithm are improved, the value of the step factor is increased at the initial stage of the firefly algorithm iteration, the search space of the firefly individual is enlarged, the optimization efficiency of the algorithm is improved, the value of the step factor is reduced at the later stage of the algorithm iteration, the convergence speed is improved, and repeated oscillation near the optimal feasible solution is effectively prevented; the improved attraction degree reduces the influence of the space distance on the attraction degree, and further improves the global optimization capability and optimization efficiency of the algorithm.
(3) In the classification and identification of the flame combustion state, the kernel parameters of the wavelet support vector machine are optimized by adopting an improved firefly algorithm, so that the method has stronger flame combustion state identification capability and faster convergence speed.
(4) The intelligent flame combustion state identification method for the fire-fighting robot extracts the global and local characteristics of flame by adopting the rotation, translation and scaling invariants of the radial Tchebichef moment, is not influenced by the position and direction of the flame in a visual field scene and the distance between the flame and image equipment, iteratively optimizes the nuclear parameters of a wavelet support vector machine based on an improved firefly algorithm, has high algorithm optimization efficiency and strong identification performance, can effectively identify and classify the combustion state of the flame, and assists the fire-fighting robot to perform subsequent targeted fire-fighting operation.
Drawings
FIG. 1 is a schematic flow chart of an intelligent flame combustion state identification method of a fire-fighting robot according to the present invention;
fig. 2 is a schematic flow chart of wavelet support vector machine kernel parameter optimization based on the improved firefly algorithm in this embodiment.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
Example 1
The embodiment provides a fire-fighting robot intelligent flame combustion state identification method, which comprises the following steps:
an image acquisition step: acquiring a series of candidate images by using a camera to construct an image data set, and obtaining a test sample set;
a segmentation pretreatment step: processing the original image of the test sample set by adopting a color criterion segmentation rule based on a YCbCr color space to obtain a corresponding segmentation processing image, then carrying out edge detection processing on the segmentation processing image by using a Canny operator, and extracting the outline of a suspected flame area to obtain an edge image;
a flame characteristic extraction step: calculating rotation, translation and scaling invariants of radial Tchebichef moments of the segmentation processing image and the edge image, respectively obtaining an area moment invariant and a contour moment invariant, and combining the area moment invariants and the contour moment invariants to construct a multi-feature composite vector for describing flame features;
and (3) flame combustion state classification and identification: and inputting the multi-feature composite vector of the test sample set into a pre-established and trained wavelet support vector machine model, predicting a classification label, and finally obtaining a classification recognition result of the flame combustion state.
As a preferred embodiment, in the segmentation preprocessing step, the expression of the color criterion segmentation rule based on the YCbCr color space is:
in the formula, I (x, y) is a segmentation processing image, if a suspected flame pixel point exists at the original image coordinate (x, y), the gray value at the corresponding coordinate is 1, otherwise, the gray value is 0; y ismeanThe average value of Y channel components of a single original image is obtained; y (x, Y) is the gray value of the brightness component of the pixel point at the coordinate (x, Y) of the original image, Cb (x, Y) is the gray value of the blue chroma component of the pixel point at the coordinate (x, Y) of the original image, and Cr (x, Y) is the gray value of the red chroma component of the pixel point at the coordinate (x, Y) of the original image; λ is an adaptation coefficient, λ > 1; s is the total number of all pixels in a single original image, (x)i,yi) Is the coordinate of the ith pixel in the original image.
As a preferred embodiment, in the flame feature extraction process, the expression of extracting flame features by using rotation, translation and scaling invariant of the radial Tchebichef moment is as follows:
in the formula (I), the compound is shown in the specification,rotation, translation and scaling invariants for the radial Tchebichef moment; t ispqA radial Tchebichef moment of order p + q; t is00A radial Tchebichef moment of zero order; n is the maximum resolution used in the angular direction; ρ (p, N) is the orthonormal squared norm;a regularized Tchebichef polynomial of order p and length N; m is N/2, and N is the side length of the image; q is a non-negative integer, and q is not more than N-1; g (r, theta) is the input image after translation and mapping; theta is 2 pi l/n, l is more than or equal to 0 and less than or equal to n-1, and l is an integer; r varies from 0 to N/2 and theta varies from 0 to 2 pi.
In a preferred embodiment, the wavelet support vector machine model is a modified firefly algorithm-wavelet support vector machine model, and the construction and training process of the modified firefly algorithm-wavelet support vector machine model is specifically,
acquiring a training sample set through the image acquisition step, and processing the training sample set through the segmentation preprocessing step and the flame feature extraction step in sequence to acquire a multi-feature composite vector of the training sample set;
initializing a pre-established firefly algorithm based on the multi-feature composite vector of the training sample set, and performing iterative search optimization on the kernel parameters of the wavelet support vector machine in the firefly transfer process by the firefly algorithm until the preset maximum iteration times are reached to obtain the optimal kernel parameters, thereby constructing an improved firefly algorithm-wavelet support vector machine model.
Further, as a preferred embodiment, the kernel function of the wavelet support vector machine adopts a Morlet wavelet function.
Further, as a preferred embodiment, the firefly algorithm adopts an improved firefly algorithm, and the improved firefly algorithm adopts an improved step factor definition formula in an iterative search optimization process to realize the self-adaptive dynamic adjustment of the search step;
the improved step-size factor definition formula enables the step-size factor to be in a variation trend of increasing first and then decreasing along with the iterative search process of the improved firefly algorithm.
Further, as a preferred embodiment, the modified step-size factor defines the expression of the formula:
in the formula (I), the compound is shown in the specification,the step size factor is the step size factor of the improved n +1 th iteration; n is the current iteration number;is the final value of the step factor;is the initial value of the step factor; n ismaxIn order to be the maximum number of iterations,is the step size factor at the nth iteration after improvement.
Further, as a preferred embodiment, in the iterative search optimization process, the improved attraction degree definition formula is adopted in the improved firefly algorithm to realize adaptive dynamic adjustment of the attraction degree;
the modified attraction degree definition formula introduces a basic attraction degree different from zero as a minimum value of the attraction degree.
Further, as a preferred embodiment, the modified attraction degree defines an expression of the formula:
in the formula (I), the compound is shown in the specification,the method is to improve the attraction degree of the firefly i to the firefly j; rho0The maximum attraction of the firefly i to other individuals; alpha is light absorption coefficient, alpha belongs to [0.1,10 ]];rijIs the Cartesian distance between firefly i and firefly j; rhobIs the basic attraction degree.
Further, as a preferred embodiment, in the iterative search optimization process, the expression of the location update formula of the firefly i is as follows:
in the formula (I), the compound is shown in the specification,is the space position coordinate vector of the firefly i at the n +1 th iteration,is the space position coordinate vector of firefly i at the nth iteration,is the space position coordinate vector of firefly j at the nth iteration, rand is in [0,1 ]]Random numbers are uniformly distributed in the interval.
The embodiment also provides an optimal implementation method, which includes the following specific implementation processes:
the flow of the fire-fighting robot intelligent flame combustion state identification method based on the radial Tchebichef moment invariant and the improved firefly algorithm-wavelet support vector machine is shown in figure 1, and the method specifically comprises the following steps:
s1: image acquisition
The method comprises the steps of collecting a series of candidate images by using a camera to construct an image data set, and dividing the collected image data set into a training sample set and a testing sample set, wherein the training sample set is used for constructing an improved firefly algorithm-wavelet support vector machine model, and the testing sample set is used for evaluating the performance of the model for recognizing the flame combustion state.
S2: segmentation pre-processing
S21: processing the original images of the training sample set and the test sample set by using a color criterion segmentation rule based on a YCbCr color space according to the following formula to obtain a segmentation processed image:
s22: in order to fully utilize the local characteristics of the flame shape subsequently, the Canny operator is used for carrying out edge detection processing on the segmentation processing image, the overall outline of the suspected flame area is extracted, an edge image is obtained, and segmentation preprocessing of the flame original image is completed.
S3: flame feature extraction
S31: calculating the radial Tchebichef moment rotation, translation and scaling invariants of each segmentation processing image and each edge image according to the following formulas to obtain corresponding area moment invariants and contour moment invariants:
s32: selecting area distance invariant and edge moment invariant with more than second order, and constructing a multi-feature composite vector F fused with global and local features:
s4: improved firefly algorithm-wavelet support vector machine construction
The wavelet support vector machine kernel parameter optimization process based on the improved firefly algorithm is shown in fig. 2, and specifically comprises the following steps:
s41: initializing an improved firefly algorithm, and setting initial values of algorithm parameters including the total number K of fireflies, the light absorption coefficient alpha and the maximum attraction rho0Final value of the step factorInitial value of step factorMaximum number of iterations nmaxAnd basic attraction degree ρb;
S42: randomly generating space position coordinate vector x of all firefly individualsn(n ═ 1,2, …, K), each spatial location coordinate vector consisting of the kernel parameters of the wavelet support vector machine;
s43: training a wavelet support vector machine to obtain a fitness function calculation value corresponding to each firefly individual so as to determine or update the corresponding absolute fluorescence brightness;
s44: sequencing all the individuals in the firefly population according to the absolute fluorescence brightness of the individuals, and taking the firefly individual with the highest absolute fluorescence brightness as the current global optimal feasible solution;
s45: evaluating the identity of each individual fireflyAbsolute fluorescent lightness of the front, and comparing the absolute fluorescent lightness L (x) of any two fireflies i and j in orderi) And L (x)j). If L (x)i)<L(xj) Firstly, the improved step-size factor is calculated according to the following formulaAnd improved attractionThen, updating the position of the firefly i according to a formula:
s46: judging whether the maximum iteration number is reached: if yes, outputting the value of the optimal kernel parameter, and performing step S47; if not, the process returns to S43.
S47: and training the wavelet support vector machine again by using the obtained optimal kernel parameters to obtain an improved firefly algorithm-wavelet support vector machine model.
S5: flame combustion state classification recognition
Inputting the flame characteristic vector of the test sample set into an improved firefly algorithm-wavelet support vector machine model, predicting a classification label of the class to which the test sample belongs, and realizing the identification of the flame combustion state; and the identification performance of the method is evaluated and analyzed by using the running time and the correct identification rate as evaluation indexes.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. An intelligent flame combustion state identification method for a fire-fighting robot is characterized by comprising the following steps:
an image acquisition step: acquiring a series of candidate images by using a camera to construct an image data set, and obtaining a test sample set;
a segmentation pretreatment step: processing the original image of the test sample set by adopting a color criterion segmentation rule based on a YCbCr color space to obtain a corresponding segmentation processing image, then carrying out edge detection processing on the segmentation processing image by using a Canny operator, and extracting the outline of a suspected flame area to obtain an edge image;
a flame characteristic extraction step: calculating rotation, translation and scaling invariants of radial Tchebichef moments of the segmentation processing image and the edge image, respectively obtaining an area moment invariant and a contour moment invariant, and combining the area moment invariants and the contour moment invariants to construct a multi-feature composite vector for describing flame features;
and (3) flame combustion state classification and identification: and inputting the multi-feature composite vector of the test sample set into a pre-established and trained wavelet support vector machine model, predicting a classification label, and finally obtaining a classification recognition result of the flame combustion state.
2. The fire-fighting robot intelligent flame burning state identification method according to claim 1, characterized in that in the segmentation preprocessing step, the expressions of the color criterion segmentation rules based on the YCbCr color space are adopted as follows:
in the formula, I (x, y) is a segmentation processing image, if a suspected flame pixel point exists at the original image coordinate (x, y), the gray value at the corresponding coordinate is 1, otherwise, the gray value is 0; y ismeanThe average value of Y channel components of a single original image is obtained; y (x, Y) is the gray value of the brightness component of the pixel point at the coordinate (x, Y) of the original image, Cb (x, Y) is the gray value of the blue chroma component of the pixel point at the coordinate (x, Y) of the original image, and Cr (x, Y) is the gray value of the red chroma component of the pixel point at the coordinate (x, Y) of the original image; λ is an adaptation coefficient, λ > 1; s is the total number of all pixels in a single original image, (x)i,yi) Is the coordinate of the ith pixel in the original image.
3. The intelligent fire-fighting robot flame combustion state identification method according to claim 1, wherein in the flame feature extraction process, an expression for extracting flame features by using rotation, translation and scaling invariants of the radial Tchebichef moment is as follows:
in the formula (I), the compound is shown in the specification,rotation, translation and scaling invariants for the radial Tchebichef moment; t ispqA radial Tchebichef moment of order p + q; t is00A radial Tchebichef moment of zero order; n is the maximum resolution used in the angular direction; ρ (p, N) is the orthonormal squared norm;a regularized Tchebichef polynomial of order p and length N; m is N/2, and N is the side length of the image; q is a non-negative integer, and q is not more than N-1; g (r, theta) is the input image after translation and mapping; theta is 2 pi l/n, l is more than or equal to 0 and less than or equal to n-1, and l is an integer; r varies from 0 to N/2 and theta varies from 0 to 2 pi.
4. The fire-fighting robot intelligent flame burning state recognition method according to claim 1, wherein the wavelet support vector machine model is an improved firefly algorithm-wavelet support vector machine model, and the construction and training process of the improved firefly algorithm-wavelet support vector machine model is specifically,
acquiring a training sample set through the image acquisition step, and processing the training sample set through the segmentation preprocessing step and the flame feature extraction step in sequence to acquire a multi-feature composite vector of the training sample set;
initializing a pre-established firefly algorithm based on the multi-feature composite vector of the training sample set, and performing iterative search optimization on the kernel parameters of the wavelet support vector machine in the firefly transfer process by the firefly algorithm until the preset maximum iteration times are reached to obtain the optimal kernel parameters, thereby constructing an improved firefly algorithm-wavelet support vector machine model.
5. The intelligent fire-fighting robot flame burning state identification method according to claim 4, wherein the kernel function of the wavelet support vector machine adopts a Morlet wavelet function.
6. The intelligent fire-fighting robot flame combustion state identification method according to claim 4, characterized in that the firefly algorithm employs an improved firefly algorithm, which employs an improved step factor definition formula in an iterative search optimization process to achieve adaptive dynamic adjustment of search step;
the improved step-size factor definition formula enables the step-size factor to be in a variation trend of increasing first and then decreasing along with the iterative search process of the improved firefly algorithm.
7. The intelligent fire fighting robot flame burning state identification method according to claim 6, wherein the improved step-size factor defines an expression of a formula as follows:
in the formula (I), the compound is shown in the specification,the step size factor is the step size factor of the improved n +1 th iteration; n is the current iteration number;is the final value of the step factor;is the initial value of the step factor; n ismaxIn order to be the maximum number of iterations,is the step size factor at the nth iteration after improvement.
8. The intelligent flame combustion state recognition method for the fire-fighting robot as claimed in claim 7, wherein the improved firefly algorithm adopts an improved attraction degree definition formula in an iterative search optimization process to realize adaptive dynamic adjustment of the attraction degree;
the modified attraction degree definition formula introduces a basic attraction degree different from zero as a minimum value of the attraction degree.
9. The intelligent fire-fighting robot flame burning state identification method according to claim 8, wherein the improved attractiveness defines an expression of a formula as follows:
in the formula (I), the compound is shown in the specification,the method is to improve the attraction degree of the firefly i to the firefly j; rho0The maximum attraction of the firefly i to other individuals; alpha is light absorption coefficient, alpha belongs to [0.1,10 ]];rijIs the Cartesian distance between firefly i and firefly j; rhobIs the basic attraction degree.
10. The intelligent flame combustion state identification method for the fire-fighting robot as recited in claim 9, wherein in the iterative search optimization process of the improved firefly algorithm, the expression of the position update formula of firefly i is as follows:
in the formula (I), the compound is shown in the specification,is the space position coordinate vector of the firefly i at the n +1 th iteration,is the space position coordinate vector of firefly i at the nth iteration,is the space position coordinate vector of firefly j at the nth iteration, rand is in [0,1 ]]Random numbers are uniformly distributed in the interval.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011010733.5A CN112052912A (en) | 2020-09-23 | 2020-09-23 | Intelligent flame combustion state identification method for fire-fighting robot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011010733.5A CN112052912A (en) | 2020-09-23 | 2020-09-23 | Intelligent flame combustion state identification method for fire-fighting robot |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112052912A true CN112052912A (en) | 2020-12-08 |
Family
ID=73604401
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011010733.5A Pending CN112052912A (en) | 2020-09-23 | 2020-09-23 | Intelligent flame combustion state identification method for fire-fighting robot |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112052912A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112560672A (en) * | 2020-12-15 | 2021-03-26 | 安徽理工大学 | Fire image recognition method based on SVM parameter optimization |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105869170A (en) * | 2016-04-13 | 2016-08-17 | 宿迁学院 | Identification and classification method for workpiece surface texture image |
CN107491756A (en) * | 2017-08-17 | 2017-12-19 | 武汉大学 | Track direction information recognition methods based on traffic sign and surface mark |
CN108229458A (en) * | 2017-12-22 | 2018-06-29 | 湖南源信光电科技股份有限公司 | A kind of intelligent flame recognition methods based on motion detection and multi-feature extraction |
-
2020
- 2020-09-23 CN CN202011010733.5A patent/CN112052912A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105869170A (en) * | 2016-04-13 | 2016-08-17 | 宿迁学院 | Identification and classification method for workpiece surface texture image |
CN107491756A (en) * | 2017-08-17 | 2017-12-19 | 武汉大学 | Track direction information recognition methods based on traffic sign and surface mark |
CN108229458A (en) * | 2017-12-22 | 2018-06-29 | 湖南源信光电科技股份有限公司 | A kind of intelligent flame recognition methods based on motion detection and multi-feature extraction |
Non-Patent Citations (3)
Title |
---|
A. EL-GHAZAL ET AL: "Scale Invariants of Radial Tchebichef Moments for Shape-Based Image Retrieval", 《2009 11TH IEEE INTERNATIONAL SYMPOSIUM ON MULTIMEDIA》 * |
MENG YANG ET AL: "Combustion State Recognition of Flame Images Using Radial Chebyshev Moment Invariants Coupled with an IFA-WSVM Model", 《APPLIED SCIENCES》 * |
吴一全 等: "基于Krawtchouk矩和支持向量机的火焰状态识别", 《中国电机工程学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112560672A (en) * | 2020-12-15 | 2021-03-26 | 安徽理工大学 | Fire image recognition method based on SVM parameter optimization |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110427825B (en) | Video flame identification method based on fusion of key frame and fast support vector machine | |
Frizzi et al. | Convolutional neural network for video fire and smoke detection | |
CN106815566B (en) | Face retrieval method based on multitask convolutional neural network | |
CN108520226B (en) | Pedestrian re-identification method based on body decomposition and significance detection | |
Zhao et al. | SVM based forest fire detection using static and dynamic features | |
CN109949278B (en) | Hyperspectral anomaly detection method based on antagonistic self-coding network | |
CN108596010B (en) | Implementation method of pedestrian re-identification system | |
Yar et al. | Optimized dual fire attention network and medium-scale fire classification benchmark | |
CN111767882A (en) | Multi-mode pedestrian detection method based on improved YOLO model | |
CN110210335B (en) | Training method, system and device for pedestrian re-recognition learning model | |
CN109255289B (en) | Cross-aging face recognition method based on unified generation model | |
CN111209818A (en) | Video individual identification method, system, equipment and readable storage medium | |
Caetano et al. | Performance evaluation of single and multiple-Gaussian models for skin color modeling | |
Bae et al. | Real-time face detection and recognition using hybrid-information extracted from face space and facial features | |
CN110728216A (en) | Unsupervised pedestrian re-identification method based on pedestrian attribute adaptive learning | |
CN113192038B (en) | Method for recognizing and monitoring abnormal smoke and fire in existing flame environment based on deep learning | |
Janku et al. | Fire detection in video stream by using simple artificial neural network | |
Chakraborty et al. | A hybrid clustering algorithm for fire detection in video and analysis with color based thresholding method | |
CN112668557A (en) | Method for defending image noise attack in pedestrian re-identification system | |
US11682111B2 (en) | Semi-supervised classification of microorganism | |
CN111274964B (en) | Detection method for analyzing water surface pollutants based on visual saliency of unmanned aerial vehicle | |
CN111898566A (en) | Attitude estimation method, attitude estimation device, electronic equipment and storage medium | |
CN112052912A (en) | Intelligent flame combustion state identification method for fire-fighting robot | |
CN108985216B (en) | Pedestrian head detection method based on multivariate logistic regression feature fusion | |
CN110796008A (en) | Early fire detection method based on video image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201208 |
|
RJ01 | Rejection of invention patent application after publication |