CN114266893A - Smoke and fire hidden danger identification method and device - Google Patents
Smoke and fire hidden danger identification method and device Download PDFInfo
- Publication number
- CN114266893A CN114266893A CN202111580662.7A CN202111580662A CN114266893A CN 114266893 A CN114266893 A CN 114266893A CN 202111580662 A CN202111580662 A CN 202111580662A CN 114266893 A CN114266893 A CN 114266893A
- Authority
- CN
- China
- Prior art keywords
- hidden danger
- image
- visible light
- thermal infrared
- primary screening
- 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
Images
Landscapes
- Image Processing (AREA)
- Fire-Detection Mechanisms (AREA)
- Radiation Pyrometers (AREA)
Abstract
A firework hidden danger identification method comprises the following steps: acquiring a thermal infrared image of a target scene; dividing and acquiring primary screening hidden danger areas in the thermal infrared image, and establishing a primary screening hidden danger area set, wherein the primary screening hidden danger area set comprises at least one primary screening hidden danger area; acquiring a visible light image of a target scene; mapping each primary screening hidden danger area in the primary screening hidden danger area set to a visible light image, and extracting a suspected hidden danger area in the visible light image, wherein the suspected hidden danger area is the mapping of the primary screening hidden danger area in the visible light image; and inputting the suspected hidden danger area into a convolutional neural network, and judging whether the suspected hidden danger area has smoke and fire hidden dangers and/or identifying the type of the smoke and fire hidden dangers. An apparatus is also provided. The invention can effectively reduce the data processing amount of the system, has higher identification and classification precision, and is particularly suitable for the field of power transmission operation and detection.
Description
Technology neighborhood
The invention belongs to the field of intelligent power transmission operation and inspection, and particularly relates to a method and equipment for identifying hidden smoke and fire hazards.
Background
Under the scene of power transmission, smoke and fire hidden dangers appear in the area under the line, and the safety of power transmission can be damaged. In recent years, with the development of deep learning technology, some firework hidden danger identification methods based on visible light images and convolutional neural networks appear, and the pressure of operation and maintenance of the power transmission line is relieved. However, it is worth noting that in some scenes, due to complexity and diversity of appearance of fireworks and smoke, and extremely high similarity between appearance characteristics of cloud and fog and night lights and appearance characteristics of fireworks and smoke, interference is brought to identification of hidden danger of fireworks, and the problems of high missing detection rate and high false alarm rate are caused.
To solve the above problems, the prior art further provides a flame detection method based on visible light and thermal imaging, such as the technical solution disclosed in chinese patent application (CN 111027541A): the flame detection method comprises the following steps: acquiring a visible light image of a site; determining the position of flame according to the visible light image by adopting a trained target detection model to obtain a first flame detection area; determining the position of flame according to the infrared thermal image by adopting a self-adaptive image segmentation method and a naive Bayes method to obtain a second flame detection area; matching calculation is carried out on the visible light image and the infrared thermal image by adopting an image registration method; and determining whether a fire alarm occurs on the site or not according to the first flame detection area, the second flame detection area and the result of the matching calculation by adopting a comprehensive decision strategy. "
Although the flame detection method proposed by the above-mentioned comparison document can reduce the missing rate and the false alarm rate, it needs to independently process each obtained visible light image and infrared thermal image, and then match and calculate the processed visible light image and infrared thermal image. For a power transmission scene with a large image processing amount, huge calculation power needs to be consumed, the operation and maintenance cost is increased, and the practicability is low.
Disclosure of Invention
The invention aims at solving the problems that in the prior art, a visible light and thermal imaging-based flame detection method needs to independently process each obtained visible light image and infrared thermal image based on a trained model respectively, and then the processed visible light image and infrared thermal image are matched and calculated, so that huge calculation power needs to be consumed, and the practicability is low in a power transmission scene with large image processing capacity.
A firework hidden danger identification method comprises the following steps: acquiring a thermal infrared image of a target scene; dividing and acquiring primary screening hidden danger areas in the thermal infrared image, and establishing a primary screening hidden danger area set, wherein the primary screening hidden danger area set comprises at least one primary screening hidden danger area; acquiring a visible light image of the target scene; mapping each primary screening hidden danger area in the primary screening hidden danger area set to the visible light image, and extracting a suspected hidden danger area in the visible light image, wherein the suspected hidden danger area is the mapping of the primary screening hidden danger area in the visible light image; and inputting the suspected hidden danger area into a convolutional neural network, and judging whether the suspected hidden danger area has smoke and fire hidden dangers and/or identifying the type of the smoke and fire hidden dangers.
A second aspect of the present invention provides a smoke and fire hazard identification device, including a first acquisition unit configured to acquire a thermal infrared image of a target scene; the segmentation unit is configured to segment and acquire primary screening hidden danger areas in the thermal infrared image and establish a primary screening hidden danger area set; the primary screening hidden danger area set comprises at least one primary screening hidden danger area; a second acquisition unit configured to acquire a visible light image of the target scene; a mapping unit, configured to map each preliminarily screened hidden danger region in the preliminarily screened hidden danger region set to the visible light image, and extract a suspected hidden danger region in the visible light image, where the suspected hidden danger region is a map of the preliminarily screened hidden danger region in the visible light image; and the convolutional neural network unit is configured to judge whether the suspected hidden danger area has the smoke and fire hidden danger or not and/or identify the type of the smoke and fire hidden danger.
Compared with the prior art, the invention has the advantages and positive effects that: on one hand, the temperature characteristic and the position characteristic in the thermal infrared image are reserved in the processing process, the visual characteristic in the visible light image is also reserved, and the missing rate and the false alarm rate are both lower than the average standard; on the other hand, the convolutional neural network only needs to process the images of the suspected hidden danger areas, and the data volume is obviously reduced. In addition, because the images of the suspected hidden danger areas are input into the convolutional neural network, more accurate pattern samples can be adopted during the training of the convolutional neural network, and the classification accuracy of the recognition of the convolutional neural network is effectively improved. The method integrally adopts a basic framework of multilevel flow, the processing speed of each step is relatively average, a plurality of steps can be executed simultaneously, the waiting time of a single step is short or basically no waiting is needed, and the method is particularly suitable for processing a large batch of graphic information.
Other features and advantages of the present invention will become more apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of an embodiment of a method for identifying a fire and smoke hazard provided by the present invention;
FIG. 2 is a flow chart of segmentation of primary screening hidden danger regions in a thermal infrared image and establishment of a hidden danger region set of the primary screening hidden danger regions;
fig. 3 is a block diagram schematically illustrating the structure of an embodiment of the smoke and fire hazard identification device provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and examples.
The terms "first," "second," "third," and the like in the description and in the claims, and in the drawings, are used for distinguishing between different objects and not necessarily for describing a particular sequential or chronological order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus.
Reference throughout this specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. One skilled in the art will appreciate that the embodiments described herein may be combined with other embodiments.
Fig. 1 depicts a flow chart of an embodiment of a method for identifying a smoke and fire hazard provided in accordance with the present invention. The method for identifying the hidden danger of the fireworks is particularly suitable for intelligent operation and detection of overhead power transmission line channels, reduces graph processing, particularly the calculation load and complexity of a convolutional neural network, and accelerates the operation speed under the condition of keeping a large amount of graph information characteristic invariance, so as to solve the problem that in the prior art, each obtained visible light image and infrared thermal image need to be independently processed, and then the processed visible light image and infrared thermal image need to be matched and calculated, so that huge calculation power needs to be consumed. The method for identifying a fire and smoke hazard of the present embodiment is described in detail below with reference to the accompanying drawings.
Step S11: acquiring a thermal infrared image of a target scene. In this embodiment, the target scene is preferably an image of the position of the overhead power transmission line channel or a position near the overhead power transmission line channel. The thermal infrared image is sampled by a thermal infrared sensor (also called infrared sensor or infrared temperature sensor). The thermal infrared image is a distribution image of the radiation temperature of the target scene, and the temperature is distinguished through different color parameters or gray levels.
Step S12: and segmenting and acquiring the primary screening hidden danger areas in the thermal infrared image, and establishing a primary screening hidden danger area set. The primary screening hidden danger areas comprise one or more primary screening hidden danger areas in a centralized mode.
Step S13: and acquiring visible light images of the same target scene. The visible light image is sampled by a visible light sensor. The visible light sensor can be an internal photoelectric effect sensor or an external photoelectric effect sensor. The type of visible light sensor or thermal infrared sensor is not limited here. The thermal infrared sensor and the visible light sensor may be provided on the same apparatus (for example, distributed vertically), or may be provided on a plurality of different apparatuses.
Step S14: unlike the prior art, the sampled visible light image is not directly input into the model for classification and identification processing in the present embodiment. Each preliminarily screened potential risk region in the preliminarily screened potential risk region set obtained in step S12 is mapped to the visible light image to extract a suspected potential risk region in the visible light image. The suspected hidden danger area is the image of the initially screened hidden danger area in the visible light image.
Step S15: inputting the suspected hidden danger area into a convolutional neural network, judging whether smoke and fire hidden dangers exist in the suspected hidden danger area or identifying the type of the smoke and fire hidden dangers, or identifying the type of the smoke and fire hidden dangers when the smoke and fire hidden dangers exist.
By the method provided by the embodiment, on one hand, the temperature characteristic and the position characteristic in the thermal infrared image and the visual characteristic in the visible light image are reserved in the processing process, and the omission ratio and the false alarm ratio are both lower than the average standard; on the other hand, the convolutional neural network only needs to process the images of the suspected hidden danger areas, and the data volume is obviously reduced. In addition, because the images of the suspected hidden danger areas are input into the convolutional neural network, more accurate pattern samples can be adopted during the training of the convolutional neural network, and the classification accuracy of the recognition of the convolutional neural network is effectively improved. The method integrally adopts a basic framework of multilevel flow, the processing speed of each step is relatively average, a plurality of steps can be executed simultaneously, the waiting time of a single step is short or basically no waiting is needed, and the method is particularly suitable for processing a large batch of graphic information.
Referring to fig. 2, a preferred process of segmenting the primary screening hidden danger areas in the thermal infrared image and establishing a hidden danger area set of the primary screening hidden danger areas will be described in detail. The method specifically comprises the following steps:
step S21: a temperature filtering threshold is set. As described above, the thermal infrared image is a distribution image of the radiation temperature of the target scene, and the temperature is distinguished by different color parameters or gray levels. Conversely, after the thermal infrared image is obtained, the radiation temperature of any point can be obtained according to the color parameter or the gray scale of any point on the thermal infrared image. The obtained radiation temperature is compared with a temperature filtering threshold value, so that whether the radiation temperature of a certain point accords with the characteristics of fire and fire hazards or not can be obtained, for example, the temperature of a fire hazard point is higher, and when the radiation temperature of a certain point is higher than the temperature filtering threshold value, the radiation temperature can be classified into one type. When the radiation temperature at a certain point is below the temperature filtering threshold, it can be classified into another class. The temperature filtering threshold is usually set to 90 degrees celsius, and may be morphologically matched color parameters or gray scale measured at 90 degrees celsius.
Step S22: and according to the fact that the radiation temperature of a certain point is higher than a temperature filtering threshold value or lower than the temperature filtering threshold value, the thermal infrared image can be binarized to obtain a thermal infrared image binary image.
The method can be specifically realized based on the following formula:
wherein IthermalFor the thermal infrared image, α0As a temperature filtering threshold, (x)i，yi) As a plane coordinate of a point in the thermal infrared image, p (x)i，yi) Is a point (x)i，yi) If the temperature filtering threshold is a color parameter or a gray scale, the temperature parameter is also converted into a corresponding color parameter or a gray scale. Thermal infrared image recording
Thermal infrared image binary mapping
B∈{0，1}，h0Is the image height, w0Is the image width; r is a set of color parameters or gray levels.
Step S23: setting a region to grow seeds s (x) in a thermal infrared image binary image0，y0)。
Specifically, traversing the thermal infrared image binary image I according to a set sequencethermal', the first value obtained is 1 as the region growing seed.
Step S24: growing seeds s (x) in regions0，y0) Starting to generate a temporary set of pixel points U on the basistempTemporary set of pixel points UtempIncluding region-grown seed s (x)0，y0)。
Step S25: region-based growth of seeds s (x)0，y0) Carrying out region growth;
performing region growing includes:
obtaining region growing seeds s (x)0，y0) Adding a point with a value of 1 to the temporary pixel point set U for the values of a plurality of pixels (8 for example) in the neighborhoodtemp；
With newly added temporary pixel point set UtempTakes the point of (1) as a reference, obtains the values of 8 pixels in the neighborhood, and adds the point with the value of 1 in the neighborhood into a temporary pixel point set Utemp(ii) a This process is repeated until a temporary pixelPoint set UtempNo longer expanded; i.e. the last one added to the temporary set of pixel points UtempAnd (4) no point with the value of 1 exists in 8 pixels in the neighborhood of the point, and the temporary pixel set is used as a primary screening hidden danger area to be added into the primary screening hidden danger area set.
Step S26: and repeating the steps S23, S24 and S25 until the thermal infrared image binary map is traversed.
Step S27: and outputting the primary screening hidden danger area set.
And when region growing seeds are set in the thermal infrared image binary image again, traversing the thermal infrared image binary image according to a set sequence by following the following process, and taking the point which has the first value of 1 and does not belong to any primary screening hidden trouble region as the region growing seeds which are set again.
In the above step S14, a specific process of mapping each primary screened hidden danger area in the primary screened hidden danger area set to the visible light image and extracting the suspected hidden danger area in the visible light image is described in detail below.
Specifically, the above steps are implemented based on the following formula:
wherein (x, y) is the plane coordinate of the midpoint of the primary screening hidden danger area, (x ', y') is the plane coordinate of the mapping point of the point in the primary screening hidden danger area in the visible light image, betasFor scaling the image scale factor, XdiffFor calibrating image offset in the abscissa dimension, YdiffThe calibrated image offset is the ordinate dimension.
In a preferred embodiment, the image scale factor β is calibratedsCalibration image offset X of abscissa dimensiondiffCalibration image offset Y in ordinate dimensiondiffAnd measuring and calculating based on a calibration process.
Specifically, the alignment of the optical axes collected by the thermal infrared sensor and the visible light sensor is performed first. A calibration reference object is placed in front of an electronic device provided with a thermal infrared sensor and a visible light sensor, and the set distance is optionally set to be 6 meters. And aligning the optical principal points of the thermal infrared sensor and the visible light sensor to the same characteristic point on the calibration reference object to complete the alignment of the optical axes. The thermal infrared sensor and the visible light sensor with the aligned optical axes or the electronic device provided with the thermal infrared sensor and the visible light sensor with the aligned optical axes are also applied to acquiring a thermal infrared image of a target scene and a visible light image of the target scene.
Setting a plurality of characteristic points on a calibration reference object, and acquiring a calibration thermal infrared image comprising all the characteristic points and a calibration visible light image comprising all the characteristic points. The calibration reference object can adopt a plate-shaped structure, and a plurality of calibration patterns are uniformly distributed on the calibration reference object. The center of each calibration pattern is used as a feature point. For example, the calibration pattern disposed on the calibration reference may be nine circles distributed in a 3 × 3 equally spaced matrix, and the centers of the nine circles are used as nine feature points. The calibration pattern on the calibration reference may also be an equilateral polygon, etc.
Calculating and calibrating image scale factor beta based on the following formulas：
Pthermal，ia coordinate point of the ith calibration characteristic point on the calibration thermal infrared image is set;
Pthermal，i-1coordinate points of the i-1 th calibration characteristic point on the calibration thermal infrared image are set;
dist(Pthermal，i，Pthermal，i-1) Calibrating the distance between two coordinate points on the thermal infrared image;
prgb，ia coordinate point of the ith calibration characteristic point on the calibration visible light image is set;
prgb，i-1a coordinate point of the i-1 th calibration characteristic point on the calibration visible light image is marked;
dist(prgb，i，prgb，i-1) For calibrating the position between two coordinate points on the visible light imageThe distance of (d);
and n is the number of line segments selected during calibration. In one embodiment, the number of line segments is the number of connecting lines between two adjacent feature points along the horizontal and vertical coordinates. Continuing with the above nine circular calibration patterns, there are 6 line segments in the abscissa direction, six line segments in the ordinate direction, and the total number of line segments is 12.
Calibration image offset X of abscissa dimensiondiffAnd a calibration image offset Y of the ordinate dimensiondiffCalculated based on the following formula:
wherein:
xrgb，ias a coordinate point prgb，iAbscissa of (a), yrgb，iAs a coordinate point prgb，iOrdinate of (a), xthermal，iAs a coordinate point Pthermal，iAbscissa of (a), ythermal，iAs a coordinate point Pthermnal，iThe ordinate of (c).
After each primary screening hidden danger area in the primary screening hidden danger area set is mapped into the visible light image, one or more suspected hidden danger areas corresponding to the primary screening hidden danger areas in the visible light image can be obtained. And inputting the suspected hidden danger area into a convolutional neural network in the form of image blocks with the same set size to judge whether the suspected hidden danger area has smoke and fire hidden dangers and identify the type of the smoke and fire hidden dangers. The convolutional neural network mainly comprises: the system comprises a feature extraction layer, a global average pooling layer and a full-connection layer, wherein the feature extraction layer is configured to extract features from a suspected hidden danger area and generate a feature map; the global average pooling layer is configured to receive the feature maps, perform dimension reduction and abstraction on the feature maps, and the full connection layer is configured to classify the feature maps after dimension reduction and abstraction, so that whether the suspected hidden danger areas have the smoke and fire hidden dangers or not is finally judged and the types of the smoke and fire hidden dangers are identified.
A second aspect of the invention provides a smoke and fire hazard identification device. As shown in fig. 3, the smoke and fire hazard identification apparatus mainly includes a first obtaining unit 11, a dividing unit 12, a second obtaining unit 13, a mapping unit 14, and a convolutional neural network unit 15.
The first acquisition unit 11 is configured to acquire a thermal infrared image of a target scene. In this embodiment, the target scene is preferably an image of the position of the overhead power transmission line channel or a position near the overhead power transmission line channel. The thermal infrared image is sampled by a thermal infrared sensor (also called infrared sensor or infrared temperature sensor). The thermal infrared image is a distribution image of the radiation temperature of the target scene, and the temperature is distinguished through different color parameters or gray levels.
The segmentation unit 12 is configured to segment and acquire the preliminary screening hidden danger regions in the thermal infrared image, and establish a preliminary screening hidden danger region set. The primary screening hidden danger area set comprises at least one primary screening hidden danger area.
The second acquisition unit 13 is configured to acquire a visible light image of the target scene. The visible light image is sampled by a visible light sensor. The visible light sensor can be an internal photoelectric effect sensor or an external photoelectric effect sensor. The type of visible light sensor or thermal infrared sensor is not limited here. The thermal infrared sensor and the visible light sensor may be provided on the same apparatus (for example, distributed vertically), or may be provided on a plurality of different apparatuses.
The mapping unit 14 is configured to map each primarily screened hidden danger region in the primarily screened hidden danger region set to the visible light image, and extract a suspected hidden danger region in the visible light image, where the suspected hidden danger region is a mapping of the primarily screened hidden danger region in the visible light image.
The convolutional neural network unit 15 is configured to determine whether a smoke and fire hazard exists in the suspected hidden hazard area, or identify the type of the smoke and fire hazard when the smoke and fire hazard exists.
On one hand, the device provided by the embodiment retains the temperature characteristics and the position characteristics in the thermal infrared image and the visual characteristics in the visible light image in the processing process, and ensures that the omission ratio and the false alarm ratio are both lower than the average standard; on the other hand, the convolutional neural network only needs to process the images of the suspected hidden danger areas, and the data volume is obviously reduced. In addition, because the images of the suspected hidden danger areas are input into the convolutional neural network, more accurate pattern samples can be adopted during the training of the convolutional neural network, and the classification accuracy of the recognition of the convolutional neural network is effectively improved. The method integrally adopts a basic framework of multilevel flow, the processing speed of each step is relatively average, a plurality of steps can be executed simultaneously, the waiting time of a single step is short or basically no waiting is needed, and the method is particularly suitable for processing a large batch of graphic information.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the technical solutions described in the foregoing embodiments, or some technical features may be substituted equally; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. A method for identifying hidden smoke and fire hazards is characterized by comprising the following steps:
acquiring a thermal infrared image of a target scene;
dividing and acquiring primary screening hidden danger areas in the thermal infrared image, and establishing a primary screening hidden danger area set, wherein the primary screening hidden danger area set comprises at least one primary screening hidden danger area;
acquiring a visible light image of the target scene;
mapping each primary screening hidden danger area in the primary screening hidden danger area set to the visible light image, and extracting a suspected hidden danger area in the visible light image, wherein the suspected hidden danger area is the mapping of the primary screening hidden danger area in the visible light image;
and inputting the suspected hidden danger area into a convolutional neural network, and judging whether the suspected hidden danger area has smoke and fire hidden dangers and/or identifying the type of the smoke and fire hidden dangers.
2. The method for identifying a pyrotechnic hazard of claim 1, wherein:
segmenting the primary screening hidden danger areas in the thermal infrared image, and establishing a hidden danger area set of the primary screening hidden danger areas comprises the following steps:
setting a temperature filtering threshold value;
utilizing the temperature filtering threshold value to binarize the thermal infrared image to obtain a thermal infrared image binary image;
setting a region growing seed in the thermal infrared image binary image;
generating a temporary set of pixel points, the temporary set of pixel points including the region growing seed;
performing region growth on the basis of the region growth seeds until the temporary pixel set is not expanded any more, and adding the temporary pixel set serving as a primary potential risk screening region into the primary potential risk screening region set;
and circularly executing from the step of setting region growing seeds in the thermal infrared image binary image until traversing the thermal infrared image binary image and outputting the primary screening hidden danger region set.
3. The method for identifying a smoke and fire hazard according to claim 2, wherein:
utilizing the temperature filtering threshold value to binarize the thermal infrared image to obtain a thermal infrared image binary image, and realizing the binary image based on the following formula:
wherein IthermalFor the thermal infrared image, α0As a temperature filtering threshold, (x)i，yi) As a plane coordinate of a point in the thermal infrared image, p (x)i，yi) Is a point (x)i，yi) The temperature parameter of (1).
4. The method for identifying a smoke and fire hazard according to claim 2, wherein:
the method for setting the region growing seeds in the thermal infrared image binary image comprises the following steps:
traversing the thermal infrared image binary image according to a set sequence, and taking the obtained point with the first value of 1 as the region growing seed;
setting a region growing seed in the thermal infrared image binary image again comprises the following steps:
traversing the thermal infrared image binary image according to a set sequence, and taking the point which has the first value of 1 and does not belong to any primary screening hidden danger area as the preset area growth seed.
5. The method for identifying a pyrotechnic potential hazard of claim 1,
mapping each primarily screened hidden danger area in the primarily screened hidden danger area set to the visible light image, and extracting suspected hidden danger areas in the visible light image based on the following formula:
wherein (x, y) is the plane coordinate of the midpoint of the primary screening hidden danger area, (x ', y') is the plane coordinate of the mapping point of the point in the primary screening hidden danger area in the visible light image, betasFor scaling the image scale factor, XdiffFor calibrating image offset in the abscissa dimension, YdiffThe calibrated image offset is the ordinate dimension.
6. The method for identifying a pyrotechnic potential hazard of claim 5,
the calibration image scale factor betasCalculated based on the following formula:
Pthermal，ia coordinate point of the ith calibration characteristic point on the calibration thermal infrared image is set;
Pthermal，i-1coordinate points of the i-1 th calibration characteristic point on the calibration thermal infrared image are set;
dist(Pthermal，i，Pthermal，i-1) Calibrating the distance between two coordinate points on the thermal infrared image;
prgb，ia coordinate point of the ith calibration characteristic point on the calibration visible light image is set;
prgb，i-1a coordinate point of the i-1 th calibration characteristic point on the calibration visible light image is marked;
dist(prgb，i，prgb，i-1) Calibrating the distance between two coordinate points on the visible light image;
and n is the number of line segments selected during calibration.
7. The method for identifying a pyrotechnic potential hazard of claim 6,
the calibration image offset X of the abscissa dimensiondiffAnd a calibration image offset Y of the ordinate dimensiondiffCalculated based on the following formula:
wherein:
xrgb，ias a coordinate point prgb，iAbscissa of (a), yrgb，iAs a coordinate point prgb，iOrdinate of (a), xthermal，iAs a coordinate point Pthermal，iAbscissa of (a), ythermal，iAs a coordinate point Pthermal，iThe ordinate of (c).
8. The method for identifying a pyrotechnic potential hazard of claim 6,
the calibration thermal infrared image and the calibration visible light image are respectively collected by a thermal infrared sensor and a visible light sensor with aligned optical axes.
9. Method for identifying a pyrotechnic potential risk according to one of claims 1 to 8,
inputting the suspected hidden danger areas in the visible light image into a convolutional neural network in the same set size; wherein the convolutional neural network comprises:
a feature extraction layer configured to extract features from the suspected potential area to generate a feature map;
the global average pooling layer is configured to receive the feature map and perform dimension reduction and abstraction on the feature map; and
a fully connected layer configured to classify the dimensionality reduced and abstracted feature map.
10. A pyrotechnic hidden danger identification device, characterized by comprising:
a first acquisition unit configured to acquire a thermal infrared image of a target scene;
the segmentation unit is configured to segment and acquire primary screening hidden danger areas in the thermal infrared image and establish a primary screening hidden danger area set; the primary screening hidden danger area set comprises at least one primary screening hidden danger area;
a second acquisition unit configured to acquire a visible light image of the target scene;
a mapping unit, configured to map each preliminarily screened hidden danger region in the preliminarily screened hidden danger region set to the visible light image, and extract a suspected hidden danger region in the visible light image, where the suspected hidden danger region is a map of the preliminarily screened hidden danger region in the visible light image; and
a convolutional neural network unit configured to determine whether a smoke and fire hazard exists in the suspected hazard area and/or identify a type of the smoke and fire hazard.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111580662.7A CN114266893A (en) | 2021-12-22 | 2021-12-22 | Smoke and fire hidden danger identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111580662.7A CN114266893A (en) | 2021-12-22 | 2021-12-22 | Smoke and fire hidden danger identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114266893A true CN114266893A (en) | 2022-04-01 |
Family
ID=80828781
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111580662.7A Pending CN114266893A (en) | 2021-12-22 | 2021-12-22 | Smoke and fire hidden danger identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114266893A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115100519A (en) * | 2022-06-23 | 2022-09-23 | 郑州儒慧信息技术有限责任公司 | Method for identifying hidden danger along high-speed rail |
| CN116052004A (en) * | 2023-02-17 | 2023-05-02 | 深圳金三立视频科技股份有限公司 | Bidirectional monitoring method and device for abnormal events, electronic equipment and storage medium |
-
2021
- 2021-12-22 CN CN202111580662.7A patent/CN114266893A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115100519A (en) * | 2022-06-23 | 2022-09-23 | 郑州儒慧信息技术有限责任公司 | Method for identifying hidden danger along high-speed rail |
| CN115100519B (en) * | 2022-06-23 | 2024-04-26 | 郑州儒慧信息技术有限责任公司 | Method for identifying hidden danger objects along high-speed rail |
| CN116052004A (en) * | 2023-02-17 | 2023-05-02 | 深圳金三立视频科技股份有限公司 | Bidirectional monitoring method and device for abnormal events, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| CN113160192B (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| CN104483326B (en) | High-voltage line defects of insulator detection method and system based on depth belief network | |
| CN109101924B (en) | Machine learning-based road traffic sign identification method | |
| CN110910378B (en) | Bimodal image visibility detection method based on depth fusion network | |
| CN116092013B (en) | Dangerous road condition identification method for intelligent monitoring | |
| CN114266893A (en) | Smoke and fire hidden danger identification method and device | |
| CN110097093A (en) | A kind of heterologous accurate matching of image method | |
| CN110889398B (en) | Multi-modal image visibility detection method based on similarity network | |
| CN106407924A (en) | Binocular road identifying and detecting method based on pavement characteristics | |
| CN106951898B (en) | Vehicle candidate area recommendation method and system and electronic equipment | |
| CN110084169B (en) | Illegal building identification method based on K-Means clustering and contour topology constraint | |
| CN107665327A (en) | A kind of method for detecting lane lines and device | |
| CN112184746A (en) | Transmission line insulator defect analysis method | |
| CN106951863B (en) | Method for detecting change of infrared image of substation equipment based on random forest | |
| CN102881160A (en) | Outdoor traffic sign identification method under low-illumination scene | |
| CN110889840A (en) | Effectiveness detection method of high-resolution 6 # remote sensing satellite data for ground object target | |
| CN112085037A (en) | Infrared thermal fault feature extraction and digital expression method for power transformation equipment | |
| Baluyan et al. | Novel approach for rooftop detection using support vector machine | |
| CN108596244A (en) | A kind of high spectrum image label noise detecting method based on spectrum angle density peaks | |
| CN113947570B (en) | Crack identification method based on machine learning algorithm and computer vision | |
| CN109579774B (en) | Antenna downward inclination angle measurement method based on depth instance segmentation network | |
| CN107705330B (en) | Visibility intelligent estimation method based on road camera | |
| CN116206208B (en) | Forestry plant diseases and insect pests rapid analysis system based on artificial intelligence | |
| CN116718599A (en) | Apparent crack length measurement method based on three-dimensional point cloud data |
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 |