WO2024045229A1

WO2024045229A1 - Early-warning blocking method and apparatus for special scenario, and electronic device and storage medium

Info

Publication number: WO2024045229A1
Application number: PCT/CN2022/119597
Authority: WO
Inventors: 马燕娟; 董志勇; 熊艳; 渠红海; 王蕾; 康谊广
Original assignee: 武汉理工光科股份有限公司
Priority date: 2022-08-29
Filing date: 2022-09-19
Publication date: 2024-03-07
Also published as: CN115457387A

Abstract

Disclosed in the present invention are an early-warning blocking method and apparatus for a special scenario, and an electronic device and a storage medium. The method comprises: acquiring a fire image to be recognized and a special scenario reference image; determining a matching feature point between said fire image and the special scenario reference image by using a preset scale-invariant feature transformation algorithm, and determining a Hausdorff distance between said fire image and the special scenario reference image according to the matching feature point; determining a size relationship between the Hausdorff distance and a preset threshold value; and if the Hausdorff distance is less than the preset threshold value, determining that said fire image is a special scenario image, and starting a fire early-warning blocking program. The present invention solves the technical problem in the prior art of repeated alarms being frequently received in special scenarios such as hot work, fire drills and facility maintenance.

Description

A special scene early warning shielding method, device, electronic equipment and storage medium

Technical field

The invention relates to the technical field of fire early warning, and specifically relates to a special scene early warning shielding method, device, electronic equipment and storage medium.

Background technique

Automatic fire alarm. In the early stage of a fire, physical quantities such as smoke, heat, and light radiation generated by combustion are converted into electrical signals through fire detectors such as temperature, smoke, and light sensors, which are transmitted to the fire alarm controller and simultaneously display the occurrence of the fire. location and record the time when the fire occurred. Automatic fire alarm systems play a very important role in building fire prevention.

With the widespread application of urban fire remote monitoring systems, the problem of the fire remote monitoring system receiving alarms from automatic fire alarm systems has also emerged: when special scenarios such as fire operations, fire drills, and regular facility maintenance occur, urban fire remote monitoring The system will frequently receive repeated alarms from a certain area of the fire host in a short period of time. Relevant personnel need to call to confirm the actual situation on site and continue to handle the alarms. This kind of repeated alarm processing process can easily cause relevant personnel to become numb to the alarm information, and other real alarm information may be determined as false alarms, causing unnecessary accidents and losses.

Therefore, it is very necessary to study a special scene early warning and shielding method by combining the Internet of Things and artificial intelligence technology.

Contents of the invention

The purpose of the present invention is to overcome the above technical deficiencies, provide a special scene early warning shielding method, device, electronic equipment and storage medium, to solve the problem of frequent repeated receiving in special scenes such as fire operations, fire drills, and facility maintenance in the existing technology. Technical issues with the alarm.

In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

In the first aspect, the present invention provides a special scene early warning shielding method, which includes:

Obtain the fire image and special scene reference image to be identified;

A preset scale-invariant feature transformation algorithm is used to determine the matching feature points between the fire image to be identified and the special scene reference image, and the fire to be identified is determined based on the matching feature points. The Hausdorff distance between the image and the reference image of the particular scene;

Determine the relationship between the Hausdorff distance and a preset threshold;

If the Hausdorff distance is less than the preset threshold, it is determined that the fire image to be identified is a special scene image, and a fire warning shielding program is started.

In some embodiments, using a preset scale-invariant feature transformation algorithm to determine the Hausdorff distance between the fire image to be identified and the special scene reference image includes:

The preset Gaussian differential function is used to determine the feature points to be identified in the fire image to be identified and the reference feature points of the special scene reference image, where the feature points to be identified and the reference feature points match each other. ;

Determine the bidirectional Hausdorff distance between the feature point to be identified and the reference feature point.

In some embodiments, determining the feature points to be identified of the fire image to be identified and the reference feature points of the special scene reference image based on a preset Gaussian differential function includes:

Based on the preset Gaussian differential function, extract the feature points to be identified of the fire image to be identified and the reference feature points of the special scene reference image;

Determine the positions of the feature points to be identified and the reference feature points, as well as the direction of the feature to be identified and the reference feature direction;

Construct a feature vector to be identified based on the position of the feature point to be identified and the direction of the feature to be identified;

Construct a reference feature vector based on the position and reference feature direction of the reference feature point;

According to the feature vector to be identified and the reference feature vector, the feature points to be identified and the reference feature points that match each other are determined.

In some embodiments, the Gaussian differential function can be expressed by the following formula:

Among them, σ is a parameter related to size, x and y represent the coordinates of image information pixels, and m and n are parameters related to feature points.

In some embodiments, determining the bidirectional Hausdorff distance between the feature point to be identified and the reference feature point includes:

Constructing a recognition set of feature points to be recognized and a reference set of reference feature points;

Traverse all feature points to be identified in the recognition set, calculate the distance between the feature points to be identified and all feature points in the reference set, determine the corresponding first shortest distance, and construct a first of the plurality of first shortest distances. target set;

Traverse the reference feature points in the reference set, calculate the distance between the reference feature point and all feature points in the identification set, determine the corresponding second shortest distance, and construct the second shortest distance of the plurality of second shortest distances. Two target sets;

Determine a first one-way Hausdorff distance according to the first target set, and determine a second one-way Hausdorff distance according to the second target set;

The larger of the first one-way Hausdorff distance and the second one-way Hausdorff distance is determined to be the two-way Hausdorff distance.

In some embodiments, obtaining a special scene reference image includes:

Obtain original images in multiple different scenarios, where the multiple different scenarios include fire work scenarios, fire drills, and regular facility maintenance;

The plurality of original images are constructed into a special scene reference image set.

In some embodiments, the scale-invariant feature transformation algorithm determines the difference value between the fire image to be identified and the special scene reference image, and further includes determining the fire image to be identified. The difference value between each original image in the special scene reference image set.

In a second aspect, the present invention also provides a special scene early warning shielding device, which includes:

The acquisition module is used to acquire fire images to be identified and special scene reference image sets;

The Hausdorff distance determination module is used to use a preset scale-invariant feature transformation algorithm to determine the matching feature points between the fire image to be identified and the special scene reference image, and determine the matching feature points based on the matching features. point, determine the Hausdorff distance between the fire image to be identified and the special scene reference image;

A judgment module used to judge the relationship between the Hausdorff distance and the threshold;

The target module, if the Hausdorff distance is less than the threshold, determines that the fire image to be identified is a special scene image, and starts a fire early warning shielding program.

In a third aspect, the present invention also provides an electronic device, including: a processor and a memory;

The memory stores a computer-readable program that can be executed by the processor;

When the processor executes the computer readable program, the steps in the special scene early warning shielding method as described above are implemented.

In a fourth aspect, the present invention also provides a computer-readable storage medium that stores one or more programs, and the one or more programs can be executed by one or more processors to Implement the steps in the special scene early warning shielding method as described above.

Compared with the existing technology, the special scene early warning shielding method, device, electronic equipment and storage medium provided by the present invention first determine the special scenes that do not require fire alarm early warning, and collect images of the special scenes to form special scene reference images. And obtain the fire image to be identified, and then use the preset scale-invariant feature transformation algorithm to determine the matching feature points between the fire image to be identified and the special scene reference image, and determine the fire to be identified by matching the feature points The Hausdorff distance between the image and the special scene reference image. Finally, by comparing the relationship between the Hausdorff distance and the preset threshold, the similarity between the fire image to be identified and the special scene reference image is determined. , determine whether the actual scene corresponding to the image to be identified is a special scene through the size of the similarity. When the Hausdorff distance is less than the preset threshold, it indicates the similarity between the fire image to be identified and the special scene reference image. Maximum, that is, the actual application scenario corresponding to the fire image to be recognized may be non-fire scenarios such as fire work and equipment maintenance, and then choose to start the shielding program so that the system no longer receives alarm information in this area.

Description of drawings

Figure 1 is a flow chart of an embodiment of the special scene early warning shielding method provided by the present invention;

Figure 2 is a flow chart of an embodiment of step S102 in the special scene early warning shielding method provided by the present invention;

Figure 3 is a flow chart of an embodiment of step S201 in the special scene early warning shielding method provided by the present invention;

Figure 4 is a flow chart of an embodiment of step S202 in the special scene early warning shielding method provided by the present invention;

Figure 5 is a schematic diagram of an embodiment of the special scene early warning shielding device provided by the present invention;

FIG. 6 is a schematic diagram of the operating environment of an embodiment of the electronic device provided by the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.

The invention relates to a special scene early warning shielding method, device, electronic equipment and storage medium, which can be applied to various industries and places, such as large warehouses, office buildings, shops, hotels, streets, etc., through the fire early warning system. The environmental conditions can be monitored in real time to keep out possible hidden dangers; currently fire detectors have different parameters for detecting fires, mainly including: temperature sensing, light sensing, smoke sensing, gas and composite detectors, and by setting fire alarms The system is connected with the public monitoring system to detect fires promptly and accurately and provide early warning, which can minimize losses.

However, in some special occasions such as hot work, equipment maintenance or fire drills, fire warnings are not required. In these situations, if too frequent fire warnings occur, it will cause fatigue and reduce the vigilance of the staff. , is not conducive to the development of fire early warning work. Therefore, it is necessary to provide a special scene early warning shielding method, device, electronic equipment and storage medium to solve the above-mentioned problems. The method, device, equipment or computer-readable storage medium involved in the present invention can be integrated with the above system or relatively independent.

Figure 1 is a flow chart of a special scene early warning shielding method provided by an embodiment of the present invention. Please refer to Figure 1. The special scene early warning shielding method includes:

S101. Obtain the fire image and special scene reference image to be identified;

S102. Use a preset scale-invariant feature transformation algorithm to determine the matching feature points between the fire image to be identified and the special scene reference image, and determine the to-be-identified feature points based on the matching feature points. Hausdorff distance between the fire image and the reference image of the special scene;

S103. Determine the relationship between the Hausdorff distance and the preset threshold;

S104. If the Hausdorff distance is smaller than the preset threshold, determine that the fire image to be identified is a special scene image, and start a fire warning shielding program.

In this embodiment, special scenes that do not require fire alarm warning are first determined, images of the special scenes are collected, a special scene reference image is formed, and the fire image to be identified is obtained, and then the preset scale-invariant feature transformation is used The algorithm determines the matching feature points between the fire image to be identified and the special scene reference image, and determines the Hausdorff distance between the fire image to be identified and the special scene reference image through the matching feature points, and finally compares The relationship between the Hausdorff distance and the preset threshold is used to determine the similarity between the fire image to be identified and the special scene reference image. Based on the similarity, it is determined whether the actual scene corresponding to the image to be identified is a special scene. , where, when the Hausdorff distance is less than the preset threshold, it means that the similarity between the fire image to be identified and the special scene reference image is the greatest, that is, the actual application scenario corresponding to the fire image to be identified may be a fire. In non-fire scenarios such as operations and equipment maintenance, you can immediately choose to start the shielding program so that the system will no longer receive alarm information in this area.

In some embodiments, please refer to Figure 2. Using a preset scale-invariant feature transformation algorithm to determine matching feature points between the fire image to be identified and the special scene reference image includes:

S201. Use the preset Gaussian differential function to determine the feature points to be identified in the fire image to be identified and the reference feature points of the special scene reference image, where the feature points to be identified and the reference feature points are in pairs. match each other;

S202. Determine the bidirectional Hausdorff distance between the feature point to be identified and the reference feature point.

In this embodiment, the SIFT algorithm (i.e., scale-invariant feature transformation algorithm) is used to extract the feature points to be identified of the fire image to be identified and the reference feature points of the special scene reference image, where the feature points to be identified are The feature points on the fire image that best reflect the actual scene represented by the image, that is, the feature points to be identified can intuitively reflect the similarity between the fire image to be identified and the special scene reference image. The reference feature points are on the special scene reference image. Feature points that best reflect the actual scene represented by the image; and calculate the two-way Hausdorff distance between the feature points to be identified and the reference feature points, and reflect the fire image and special scene to be identified through the Hausdorff distance The similarity between the reference images is used to determine the fire situation at the fire location to be identified, thereby guiding the work of the fire alarm.

In some embodiments, please refer to Figure 3. Based on the preset Gaussian differential function, determining the feature points to be identified in the fire image to be identified and the reference feature points of the special scene reference image include:

S301. Based on the preset Gaussian differential function, extract the feature points to be identified of the fire image to be identified and the reference feature points of the special scene reference image;

S302. Determine the positions of the feature points to be identified and the reference feature points, as well as the direction of the feature to be identified and the direction of the reference feature;

S303. Construct a feature vector to be identified based on the position of the feature point to be identified and the direction of the feature to be identified;

S304. Construct a reference feature vector based on the position and reference feature direction of the reference feature point;

S305. According to the feature vector to be identified and the reference feature vector, determine the feature points to be identified and the reference feature points that match each other.

In this embodiment, the feature points to be identified and the reference feature points are extracted in order to search for image positions on all scale spaces, and to identify potential scale- and rotation-invariant interest points through a preset Gaussian differential function, where , the preset Gaussian differential function can be expressed by the following formula:

Among them, σ is a parameter related to size, x and y represent the coordinates of image information pixels, m and n are parameters related to feature points. For example, m and n can be randomly set parameter values, which are consistent with Gaussian differential function parameters. Define requirements.

It should be noted that at each candidate position, a fine-fitting model is used to determine the position and scale of the feature point to be identified and the reference feature point, and then the position is assigned to each feature point based on the local gradient direction of the image. one or more directions;

Further, according to the position and direction of the feature points, the feature vector to be identified and the reference feature vector are constructed respectively. Finally, by comparing the feature points in the feature vector to be identified and the reference feature vector, a number of feature points to be identified that match each other are found. and reference feature points.

In some embodiments, referring to Figure 4, determining the bidirectional Hausdorff distance between the feature point to be identified and the reference feature point includes:

S401. Construct a recognition set of feature points to be recognized and a reference set of reference feature points;

S402. Traverse all feature points to be identified in the recognition set, calculate the distance between the feature points to be identified and all feature points in the reference set, determine the corresponding first shortest distance, and construct a plurality of first shortest distances. First target set;

S403. Traverse the reference feature points in the reference set, calculate the distance between the reference feature point and all feature points in the recognition set, determine the corresponding second shortest distance, and construct the multiple second shortest distances. the second target set;

S404. Determine the first one-way Hausdorff distance according to the first target set, and determine the second one-way Hausdorff distance according to the second target set;

S405. Determine whichever of the first one-way Hausdorff distance and the second one-way Hausdorff distance is larger is the two-way Hausdorff distance.

In this embodiment, the recognition set and the reference set are respectively constructed based on the feature points to be recognized and the reference feature points that match each other in the feature vector to be recognized and the reference feature vector. Specifically, the feature points to be recognized in the recognition set are Points that can match the reference feature points. The reference feature points in the reference set are points that can match the feature points to be identified.

In a specific embodiment, assuming that the recognition set is A={a1, a2,...,an} and the reference set is B={b1, b2,...,bn}, the process of determining the bidirectional Hausdorff distance is for the following steps:

First, take a point a1 in set A, calculate the distance from a1 to all points in set B, and retain the shortest distance d1;

Then traverse all points in the set A and calculate d2,...,dn;

Then compare all distances {d1, d2,...,dn}, and select the longest distance dx, which is the one-way Hausdorff distance of A→B, recorded as h(A, B);

Then follow the above steps to calculate the one-way Hausdorff distance of B→A, recorded as h(B, A);

Finally, the longest distance between h(A, B) and h(B, A) is selected, which is the two-way Hausdorff distance of the set A and B.

In some embodiments, obtaining a special scene reference image further includes:

In this embodiment, for different special scenarios, including but limited to fire operations, equipment maintenance or fire drills, corresponding sample images need to be collected to form a special scene reference image set to avoid missed selections during the actual discrimination process. Condition.

In this embodiment, the fire image to be identified and the special scene reference image set are compared for similarity until the most similar picture is found or the comparison is completed with all pictures.

Based on the above special scene early warning shielding method, the embodiment of the present invention also provides a special scene early warning shielding device 500. Please refer to Figure 5. The special scene early warning shielding device 500 includes an acquisition module 510 and a Hausdorff distance determination module 520. , judgment module 530 and target module 540.

The acquisition module 510 is used to acquire the fire image and special scene reference image set to be identified;

The Hausdorff distance determination module 520 is used to use a preset scale-invariant feature transformation algorithm to determine the matching feature points between the fire image to be identified and the special scene reference image, and based on the matching Feature points determine the Hausdorff distance between the fire image to be identified and the special scene reference image;

The judgment module 530 is used to judge the relationship between the Hausdorff distance and the threshold;

Target module 540, if the Hausdorff distance is less than the threshold, determines that the fire image to be identified is a special scene image, and starts a fire warning shielding program.

As shown in Figure 6, based on the above special scene early warning shielding method, the present invention also provides an electronic device. The electronic device can be a mobile terminal, a desktop computer, a notebook, a palmtop computer, a server and other computing devices. The electronic device includes a processor 610, a memory 620 and a display 630. FIG. 6 only shows some components of the electronic device, but it should be understood that implementation of all the components shown is not required, and more or fewer components may be implemented instead.

In some embodiments, the memory 620 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. In other embodiments, the memory 620 may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), or a secure digital (Secure Digital, SD) card equipped on the electronic device. Flash Card, etc. Further, the memory 620 may also include both an internal storage unit of the electronic device and an external storage device. The memory 620 is used to store application software and various data installed on the electronic device, such as program codes for installing the electronic device. The memory 620 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a special scene early warning shielding program 640 is stored in the memory 620, and the special scene early warning shielding program 640 can be executed by the processor 610, thereby realizing the special scene early warning shielding method in each embodiment of the present application.

In some embodiments, the processor 610 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code stored in the memory 620 or process data, such as executing special scene warnings. Shielding methods, etc.

In some embodiments, the display 630 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. The display 630 is used to display information on the special scene early warning shielding device and to display a visual user interface. The components of the electronic device 610-630 communicate with each other via the system bus.

Of course, those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware (such as processors, controllers, etc.) through computer programs. The programs can be stored in a computer. In a computer-readable storage medium, when executed, the program may include the processes of the above method embodiments. The storage medium may be a memory, a magnetic disk, an optical disk, etc.

The above-described specific embodiments of the present invention do not limit the scope of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

A special scene early warning shielding method is characterized by including:

Obtain the fire image and special scene reference image to be identified;

A preset scale-invariant feature transformation algorithm is used to determine the matching feature points between the fire image to be identified and the special scene reference image, and the fire to be identified is determined based on the matching feature points. The Hausdorff distance between the image and the reference image of the particular scene;

Determine the relationship between the Hausdorff distance and a preset threshold;

If the Hausdorff distance is less than the preset threshold, it is determined that the fire image to be identified is a special scene image, and a fire warning shielding program is started.
The special scene early warning shielding method according to claim 1, characterized in that the preset scale-invariant feature transformation algorithm is used to determine the match between the fire image to be identified and the special scene reference image. Feature points include:

The preset Gaussian differential function is used to determine the feature points to be identified in the fire image to be identified and the reference feature points of the special scene reference image, where the feature points to be identified and the reference feature points match each other. .
The special scene early warning shielding method according to claim 2, characterized in that the preset Gaussian differential function is used to determine the feature points of the fire image to be identified and the reference feature points of the special scene reference image. ,include:

Based on the preset Gaussian differential function, extract the feature points to be identified of the fire image to be identified and the reference feature points of the special scene reference image;

Determine the positions of the feature points to be identified and the reference feature points, as well as the direction of the feature to be identified and the reference feature direction;

Construct a feature vector to be identified based on the position of the feature point to be identified and the direction of the feature to be identified;

Construct a reference feature vector based on the position and reference feature direction of the reference feature point;

According to the feature vector to be identified and the reference feature vector, the feature points to be identified and the reference feature points that match each other are determined.
The special scene early warning shielding method according to claim 3, characterized in that the Gaussian differential function can be expressed by the following formula:

Among them, σ is a parameter related to size, x and y represent the coordinates of image information pixels, and m and n are parameters related to feature points.
The special scene early warning shielding method according to claim 2, wherein the Hausdorff difference between the fire image to be identified and the special scene reference image is determined based on the matching feature points. distance, including:

Constructing a recognition set of feature points to be recognized and a reference set of reference feature points;

Traverse all feature points to be identified in the recognition set, calculate the distance between the feature points to be identified and all feature points in the reference set, determine the corresponding first shortest distance, and construct a first of the plurality of first shortest distances. target set;

Traverse the reference feature points in the reference set, calculate the distance between the reference feature point and all feature points in the identification set, determine the corresponding second shortest distance, and construct the second shortest distance of the plurality of second shortest distances. Two target sets;

Determine a first one-way Hausdorff distance according to the first target set, and determine a second one-way Hausdorff distance according to the second target set;

The larger of the first one-way Hausdorff distance and the second one-way Hausdorff distance is determined to be the two-way Hausdorff distance.
The special scene early warning shielding method according to claim 1, characterized in that said obtaining a special scene reference image further includes:

Obtain original images in multiple different scenarios, where the multiple different scenarios include fire work scenarios, fire drills, and regular facility maintenance;

The plurality of original images are constructed into a special scene reference image set.
The special scene early warning shielding method according to claim 6, characterized in that the scale-invariant feature transformation algorithm determines the difference value between the fire image to be identified and the special scene reference image, and further The method includes determining a difference value between the fire image to be identified and each original image in the special scene reference image set.
A special scene early warning shielding device, which is characterized by including:

The acquisition module is used to acquire fire images to be identified and special scene reference image sets;

The Hausdorff distance determination module is used to use a preset scale-invariant feature transformation algorithm to determine the matching feature points between the fire image to be identified and the special scene reference image, and determine the matching feature points based on the matching features. point, determine the Hausdorff distance between the fire image to be identified and the special scene reference image;

A judgment module used to judge the relationship between the Hausdorff distance and the threshold;

The target module, if the Hausdorff distance is less than the threshold, determines that the fire image to be identified is a special scene image, and starts a fire warning shielding program.
An electronic device, characterized by including: a processor and a memory;

The memory stores a computer-readable program that can be executed by the processor;

When the processor executes the computer readable program, the steps in the special scene early warning shielding method as described above are implemented.
A computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned The steps in the special scene early warning shielding method.