A method, device and storage medium for detecting fireproof clothing

Abstract

The present application discloses a method, device and storage medium for detecting fireproof clothing, which are used to improve the accuracy of detection. The present application includes: collecting behavioral videos of staff at at least two fireproof clothing detection points, each of which includes at least one acquisition device; obtaining at least one image to be detected of the same target staff from the behavioral video; determining the brightness data of the image to be detected; when the brightness data does not meet the preset brightness condition, enhancing the image to be detected; inputting at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of fireproof clothing, and outputting at least one detection data; and alerting the target staff according to the detection data.

Description

Technical Field

[0001] The present application relates to the field of target detection, and in particular to a method, device and storage medium for detecting fireproof clothing. Background Art

[0002] Fireproof clothing is an essential piece of clothing for rescue workers and industrial workers at work. It is designed to effectively protect the wearer's personal safety from flames, high temperatures and other related dangers. It combines multiple functions such as fireproofing and heat insulation, protection against harmful gases, and resistance to wear and tear, providing comprehensive protection for the wearer. Wearing fireproof clothing is relatively complicated, which may lead to omissions when the wearer wears it.

[0003] In the prior art, in a dimly lit environment, a camera is used to capture video images to detect whether the fireproof clothing is worn completely. The averaging of image brightness and the loss of image details caused by low light conditions will lead to a decrease in detection accuracy. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a method, device and storage medium for detecting fire retardant clothing.

[0005] The technical solution provided in this application is described below:

[0006] The first aspect of the present application provides a method for detecting fire retardant clothing, the method comprising:

[0007] Collecting worker behavior videos at at least two fire-resistant clothing inspection points, each fire-resistant clothing inspection point including at least one collection device;

[0008] Obtain at least one image to be detected of the same target worker from the behavior video;

[0009] Determine brightness data of the image to be detected;

[0010] When the brightness data does not meet the preset brightness conditions, the image to be detected is enhanced;

[0011] Inputting at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and outputting at least one detection data;

[0012] Alert the target staff based on the detection data.

[0013] Optionally, after obtaining at least one image to be detected of the same target worker from the behavior video and before determining the brightness data of the image to be detected, the detection method further includes:

[0014] Image data of the same scene collected by the same camera device is selected, and the frame difference method is used to process the two frames of images. The foreground image area retained after the frame difference method is used to determine whether there is a target in the image, and the shape and area of ​​the foreground target area are used to set a threshold for background interference filtering.

[0015] The fireproof clothing feature detection area is determined on the image to be detected through the reserved foreground area, and the fireproof clothing feature detection area includes a fireproof helmet detection area, a fireproof clothing detection area, a fireproof pants detection area, a boots detection area, a glove detection area and a belt detection area.

[0016] Optionally, determining brightness data of the image to be detected includes:

[0017] Determine the initial background area of ​​the image to be detected through the fireproof clothing feature detection area;

[0018] Screening out the staff area other than the target staff in the initial background area to generate a target background area;

[0019] Calculate the grayscale mean of the image to be detected in the target background area, and determine the grayscale mean as brightness data.

[0020] Optionally, each component of the fireproof clothing is provided with a reflective area, and the reflective shape characteristics of the reflective area of ​​each component are different;

[0021] Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data, including:

[0022] Input at least one image to be detected into a pre-trained convolutional neural network model;

[0023] The convolutional neural network model is used to identify the reflective shape features of each fire retardant clothing feature detection area in at least one image to be detected, and at least one detection data is generated. The detection data includes the recognition results of specific reflective shape features in each fire retardant clothing feature detection area of ​​one image to be detected.

[0024] Optionally, at least one image to be detected of the same target worker is obtained from the behavior video, including:

[0025] uniformly collecting at least two initialization images from the behavior video, and selecting foreground area images of the same target worker as a first image set from the at least two initialization images by using a frame difference method;

[0026] Detecting the target staff area of ​​each image in the first image set, and filtering out images that do not meet the area requirements, so as to generate at least two images to be collected;

[0027] Optionally, when the brightness data does not meet the preset brightness condition, the image to be detected is enhanced, including:

[0028] When the brightness data does not meet the preset brightness conditions, the CSNorn model is used to enhance the fireproof clothing feature detection area of ​​the image to be detected.

[0029] Optionally, when the brightness data does not meet the preset brightness condition, after the image to be detected is enhanced, before at least one image to be detected is input into a pre-trained convolutional neural network model for fire protection clothing feature detection and before at least one detection data is output, the detection method includes:

[0030] The image to be detected is randomly cropped and segmented, and the segmented image is geometrically compressed and filled with zeros to ensure that the target of the image to be detected does not produce excessive deformation and to ensure that small targets are not missed due to image compression.

[0031] A second aspect of the present application provides a detection device for fire retardant clothing, the method comprising:

[0032] A collection unit, used to collect behavior videos of workers at at least two fire protection clothing detection points, each of which includes at least one collection device;

[0033] An acquisition unit, used for acquiring at least one to-be-detected image of the same target worker from the behavior video;

[0034] A determination unit, used to determine the brightness data of the image to be detected;

[0035] A processing unit, configured to perform enhancement processing on the image to be detected when the brightness data does not meet a preset brightness condition;

[0036] An output unit, used for inputting at least one of the images to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and outputting at least one detection data;

[0037] An alarm unit is used to alarm the target staff according to the detection data.

[0038] A third aspect of the present application provides a detection device for fire retardant clothing, the device comprising:

[0039] Processor, memory, input-output unit, and bus;

[0040] The processor is connected to the memory, the input and output unit, and the bus;

[0041] The memory stores a program, and the processor calls the program to execute the first aspect and the optional detection method of any one of the first aspect.

[0042] A fourth aspect of the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed on a computer, the program executes the first aspect and any optional detection method in the first aspect.

[0043] It can be seen from the above technical solutions that this application has the following advantages:

[0044] 1. The accuracy of fire retardant clothing wearing complete detection is improved by determining whether the image needs to be adjusted for brightness.

[0045] 2. By enhancing the brightness of the video image, it is possible to more clearly and quickly determine whether the fireproof clothing is worn completely. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution in the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 A schematic diagram of an embodiment of the fire retardant clothing detection method of the present application;

[0048] Figure 2 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0049] Figure 3 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0050] Figure 4 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0051] Figure 5 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0052] Figure 6 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0053] Figure 7 This is a schematic diagram of another embodiment of the fire retardant clothing detection method of the present application;

[0054] Figure 8 A schematic diagram of an embodiment of a processing device of the fire retardant clothing detection method of the present application;

[0055] Fig. 9 A schematic diagram of another embodiment of a processing device for the fire retardant clothing detection method of the present application. DETAILED DESCRIPTION

[0056] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0057] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0058] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0059] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0060] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0061] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0062] In the prior art, fireproof clothing is an essential clothing equipment for personnel engaged in safety rescue and industrial workers at work, which is designed to effectively protect the personal safety of the wearer from flames, high temperatures and other related dangers. Its multiple functions such as fireproofing and heat insulation, protection against harmful gases, wear resistance and tear resistance provide comprehensive protection for the wearer. The wearing of fireproof clothing is relatively complicated, which may lead to omissions when the wearer wears it.

[0063] In the prior art, in a dimly lit environment, a camera is used to capture video images to detect whether the fireproof clothing is worn completely. The averaging of image brightness and the loss of image details caused by low light conditions will lead to a decrease in detection accuracy.

[0064] Based on this, the present application discloses a method, device and storage medium for detecting fire retardant clothing, which are used to improve the accuracy of detection.

[0065] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0066] The method of the present application can be applied to a server, a device, a terminal or other devices with logic processing capabilities, and the present application does not limit this. For the convenience of description, the following description is made by taking the execution subject as an example of a terminal.

[0067] See also Figure 1 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0068] 101. Collecting the behavior videos of the staff at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0069] The fire-resistant clothing detection points in this embodiment are set at the entrance of the detection point and the area where the fire-resistant clothing is worn. Each detection point is equipped with 4 cameras, which respectively collect images from the front, back, left and right directions of the staff, which helps to achieve more comprehensive monitoring coverage, reduce monitoring blind spots, and improve the efficiency and accuracy of safety management. Through the collaborative work of multiple points, the behavioral dynamics of the staff can be more comprehensively grasped. At each fire-resistant clothing detection point, at least one acquisition device is equipped to capture and record the behavior video of the staff wearing fire-resistant clothing in the area in real time. Through video playback, the specific behavior of the staff at a specific time and place can be clearly seen. The above case is only one example.

[0070] 102. Obtain at least one image to be detected of the same target staff member from the behavior video;

[0071] Use a video processing library (such as OpenCV) to read the video file and parse it frame by frame. Use a target detection algorithm (such as YOLO, SSD, Faster R-CNN, etc.) to process each frame and identify the target staff in the video. The accuracy of detection can be improved based on features such as work clothes color and identity. The above case is just one example.

[0072] 103. Determine brightness data of the image to be detected;

[0073] In this embodiment, statistics such as average brightness, maximum brightness, minimum brightness, etc. can be calculated, or histogram analysis can be performed to understand the brightness distribution of the image. A brightness threshold is then set according to the brightness required for the detected image. When the brightness of the image to be detected does not reach this threshold, the brightness of the image to be detected is adjusted. Determining a fixed threshold is more convenient and quick for judging whether the brightness data of the image to be detected meets the conditions for detection. The above case is only one example.

[0074] 104. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0075] In this embodiment, the brightness of the image is first detected before the detection image is enhanced. Traditional target detection algorithms are significantly limited by changes in lighting conditions, especially in nighttime environments. Since the brightness of the image drops significantly and the target features become blurred, the detection accuracy and recognition efficiency of the model are greatly reduced, posing a significant challenge to the effective detection of the target.

[0076] Therefore, when the acquired brightness data is judged not to meet the preset brightness conditions, an image processing tool (such as the CSNorm model) is used to enhance the image to be detected. Specifically, the overall brightness of the image is improved from aspects such as automatic brightness adjustment, contrast enhancement, local area processing, and color correction to make it more suitable for subsequent processing; to ensure the significance of important features, thereby improving the accuracy of subsequent analysis. The steps of the CSNorm model to enhance the image are as follows:

[0077] Automatic brightness adjustment: The CSNorm tool will first automatically identify the brightness level of the image and adjust the brightness of the image according to the preset conditions set by the user to ensure that the brightness of the image meets a certain standard.

[0078] Contrast enhancement: After brightness adjustment, use the contrast enhancement function of CSNorm to improve the contrast of the image and make the important features in the image more prominent.

[0079] Local area processing: Adjust the brightness and contrast of specific areas in the image as needed to highlight important information or details.

[0080] Color Correction: Finally, the color correction function of CSNorm is used to adjust the color channels of the image to ensure that the color of the image is not distorted during the enhancement process while enhancing the visual effect.

[0081] 105. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0082] In this embodiment, a convolutional neural network model is used to perform feature detection on fireproof clothing. The convolutional neural network model here refers to YOLOv5 (You Only Look Once version 5), which is a real-time target detection algorithm based on deep learning. YOLOv5 regards the target detection task as a regression problem and predicts the category and position of the target directly from the image. The core idea is to divide the image into multiple grids, each of which is responsible for predicting the target within the grid. This single detection method allows the entire image to obtain the position and category of all objects through only one forward pass.

[0083] YOLOv5 can achieve fast target detection while maintaining high accuracy. On the GPU, YOLOv5 can achieve a very high processing speed (such as 140FPS), which makes it very suitable for real-time applications. In addition, it uses advanced anchor box prediction methods and feature fusion technology to improve the accuracy and robustness of detection.

[0084] In image processing, when the same object is marked by multiple target boxes at the same time, the intersection-of-union ratio can be used as a measure of the overlap between these target boxes. IOU quantifies the degree of overlap by calculating the ratio of the area of ​​the intersection of two target boxes to the area of ​​their union. The higher this ratio is, the greater the overlap between the two target boxes.

[0085] Based on the intersection and the union, determining an intersection-and-union ratio of two adjacent video images;

[0086] In the embodiment of the present invention, the intersection-and-union ratio of two adjacent video images may be calculated by the following formula:

[0087]

[0088] In the formula, IOU is the intersection-over-union ratio of the target box, A is one of the target boxes, and B is any target box except A.

[0089] It can be understood that the higher the target frame overlap, the closer the IOU should be to 1. When the target frame overlap is low, the intersection of the target frames will be smaller, and the IOU should be closer to 0. Therefore, when the intersection-union ratio is the largest, it can be considered to be the same target.

[0090] 106. Alert the target staff based on the detection data.

[0091] Alerting target workers based on detection data is a comprehensive process involving real-time data processing, target identification and tracking, and alarm notification. This process is often used in many fields such as safety monitoring, workplace safety, and health monitoring.

[0092] In this embodiment, the alarm is used in the field of security monitoring. The target staff is warned of unqualified behavior, and the alarm method can be a sound alarm, SMS notification, email reminder, and APP push.

[0093] After alerting the target staff, the follow-up process records the detailed information of each alert, including time, location, reason, and processing results, and generates a report for subsequent analysis. Based on the alarm processing results and staff feedback, the detection algorithm and alarm strategy are continuously optimized to improve the accuracy and efficiency of the system.

[0094] See also Figure 2 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0095] 201. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0096] 202. Obtain at least one to-be-detected image of the same target worker from the behavior video;

[0097] 203. Select image data of the same scene collected by the same camera device, use the frame difference method to process the two frames of images, determine whether there is a target in the image through the foreground image area retained after the frame difference method, and set a threshold to perform background interference filtering according to the shape and area of ​​the foreground target area;

[0098] In this embodiment, two consecutive frames of images are first read, the two frames are converted into grayscale images, and then their differences are calculated. By finding the contour, the foreground target can be identified and filtered according to its shape and area. The frame difference method works better for targets captured from behavioral videos, and the threshold value needs to be adjusted according to the actual application scenario.

[0099] 204. Determine a fireproof clothing feature detection area on the image to be detected by using the reserved foreground area, wherein the fireproof clothing feature detection area includes a fireproof helmet detection area, a fireproof clothing detection area, a fireproof pants detection area, a boots detection area, a glove detection area, and a belt detection area;

[0100] In this embodiment, the human body structure area on the target human body frame is first determined, and the image to be detected is distinguished into the human body structure part and the background part. Secondly, the image to be detected is identified according to the human body structure part, and the image to be detected with a complete human body structure is screened out. Finally, the human body structure part on the target human body frame is determined as the fireproof clothing feature detection area, and the monitoring area is divided into six areas: helmet detection area, top detection area, bottom detection area, boots detection area, glove detection area and belt detection area. Specifically, reflective strips of different shapes are sewn on the fireproof clothing in each different area, and the reflective strips are sewn in a circle in each different area to ensure that the target staff can be identified in 360° direction to see whether the fireproof clothing is fully worn, thereby reducing accidents caused by blind spots. The above case is only one example.

[0101] 205. Determine brightness data of the image to be detected;

[0102] 206. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0103] 207. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0104] 208. Alert the target staff based on the detection data.

[0105] In this embodiment, steps S201 to S202 are implemented in a similar manner to steps S101 to S102 in the above embodiment, and are not described in detail here. Steps S205 to S208 are similar to steps S103 to S106 in the above embodiment, and are not described in detail here.

[0106] See also Figure 3 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0107] 301. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0108] 302. Acquire at least one to-be-detected image of the same target staff member from the behavior video;

[0109] 303. Determine the initial background area of ​​the image to be detected through the fire protection clothing feature detection area;

[0110] 304. Screening out the staff area other than the target staff in the initial background area to generate a target background area;

[0111] 305. Calculate the grayscale mean of the image to be detected in the target background area, and determine the grayscale mean as brightness data.

[0112] In this embodiment, the target background area is determined by threshold segmentation, and the target background area is extracted from the image according to the position determined by segmentation. In image processing, Python's OpenCV library can be used to read the image to be detected and calculate the grayscale mean of the image. First, if the image is colored (i.e., it has three color channels BGR), it needs to be converted into a grayscale image, because the grayscale image has only one channel, and the brightness value of each pixel ranges from 0 (black) to 255 (white). Then the average of the grayscale values ​​of all pixels is calculated.

[0113] The mean of the grayscale values ​​is used in conjunction with other statistics (such as the grayscale standard deviation) to assess the contrast of an image. Based on the mean grayscale value, the image is automatically adjusted to enhance its visual effect. For example, if the mean grayscale value is low, the brightness of the image can be automatically increased.

[0114] 306. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0115] 307. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0116] 308. Alert the target staff based on the detection data.

[0117] In this embodiment, steps S301 to S302 are implemented in a similar manner to steps S101 to S102 in the above embodiment, and are not described in detail here. Steps S306 to S308 are similar to steps S104 to S106 in the above embodiment, and are not described in detail here.

[0118] See also Figure 4 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0119] 401. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0120] 402. Acquire at least one to-be-detected image of the same target worker from the behavior video;

[0121] 403. Determine brightness data of the image to be detected;

[0122] 404. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0123] 405. Input at least one of the to-be-detected images into a pre-trained convolutional neural network model;

[0124] 406. Recognize the reflective shape features of each fire retardant clothing feature detection area in at least one of the images to be detected by using the convolutional neural network model to generate at least one detection data, wherein the detection data includes a recognition result of a specific reflective shape feature in each fire retardant clothing feature detection area of ​​the image to be detected;

[0125] In this embodiment, the convolutional neural network model uses its feature representation ability learned in the training stage to automatically locate the fireproof clothing feature detection area in the input image to be detected. For each located fireproof clothing feature detection area, the model further performs fine feature extraction and classification tasks. Through multi-layer convolution, the model is able to capture the unique features of the reflective shape in the area. Subsequently, the classification layer is used to map these features to predefined reflective shape feature categories to generate recognition results for each area. For each image to be detected, the model will generate one or more detection data items. Each detection data item records in detail the recognition results of a certain fireproof clothing feature detection area in the image for a specific reflective shape feature, including but not limited to the type of reflective shape, position coordinates, confidence score and other information. Finally, the model outputs the detection data in an appropriate format for use by users or subsequent processing systems. At the same time, the accuracy of the detection results can also be evaluated, and the performance and generalization ability of the model can be continuously optimized by comparing with actual labeled data or expert judgment.

[0126] 407. Alert the target staff member based on the detection data.

[0127] In this embodiment, steps S401 to S405 are implemented in a similar manner to steps S101 to S105 in the above embodiment, and are not described in detail here. Step S407 is similar to step S106 in the above embodiment, and is not described in detail here.

[0128] See also Figure 5 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0129] 501. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0130] 502. uniformly collect at least two initialization images from the behavior video, and select foreground area images of the same target worker as a first image set from the at least two initialization images by using a frame difference method;

[0131] Frame Difference Method is a commonly used video processing technology, which is used to detect the changes between adjacent frames in a video sequence and identify moving objects. This method is more suitable for scenes such as surveillance video analysis and target tracking.

[0132] The frame difference method is mainly based on the time difference of pixels. When there are moving objects in the video, there will be obvious differences in grayscale or brightness between adjacent frames. By performing differential operations on these adjacent frames, a difference image between two frames can be obtained. In this difference image, the moving object part will show a higher brightness or grayscale value, while the static background part will be relatively dim. In this way, the foreground area image to be used can be better distinguished and obtained.

[0133] 503. Detect the target staff area of ​​each image in the first image set, and filter out images with unqualified area, so as to generate at least two images to be collected;

[0134] The sampling interval is calculated based on the total number of frames in the video and the expected number of initialization images (at least two). Based on the calculated sampling interval, at least two initialization images are evenly extracted from the video. For each initialization image, the target detection algorithm is used to identify all the workers in the image, and feature matching is performed based on each target feature. Based on the result of feature matching, images belonging to the same target worker are screened out.

[0135] If there is uncertainty, further manual confirmation or the use of more complex tracking algorithms is required to assist in the judgment.

[0136] 504. Determine brightness data of the image to be detected;

[0137] 505. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0138] 506. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0139] 507. Alert the target staff based on the detection data.

[0140] In this embodiment, step S501 is implemented in a similar manner to step S101 in the above embodiment, and will not be described in detail here. Steps S504 to S507 are similar to steps S103 to S106 in the above embodiment, and will not be described in detail here.

[0141] See also Figure 6 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0142] 601. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0143] 602. Acquire at least one to-be-detected image of the same target worker from the behavior video;

[0144] 603. Determine brightness data of the image to be detected;

[0145] 604. When the brightness data does not meet the preset brightness condition, the CSNorm model is used to enhance the fireproof clothing feature detection area of ​​the image to be detected;

[0146] Conventional target detection algorithms are greatly affected by lighting. Especially at night, low brightness and unclear target features have a great impact on the detection rate of the model, making it impossible to determine the designated boundary between the target worker and the background. CSNorm is a technology that can improve the model's lighting adaptability. It can generalize unknown lighting conditions while maintaining the reconstruction capability of known lighting conditions, and enhance the human frame outline in the image, thereby improving the model's image detection rate when the brightness conditions are insufficient.

[0147] 605. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0148] 606. Alert the target staff member based on the detection data.

[0149] In this embodiment, steps S601 to S603 are implemented in a similar manner to steps S101 to S103 in the above embodiment, and are not described in detail here. Steps S605 to S606 are similar to steps S105 to S106 in the above embodiment, and are not described in detail here.

[0150] See also Figure 7 The present application provides an embodiment of a method for detecting fire retardant clothing, comprising:

[0151] 701. Collecting behavior videos of staff members at at least two fire protection clothing inspection points, each fire protection clothing inspection point including at least one collection device;

[0152] 702. Acquire at least one to-be-detected image of the same target worker from the behavior video;

[0153] 703. Determine brightness data of the image to be detected;

[0154] 704. When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced;

[0155] 705. Randomly crop and split the image to be detected, and perform geometric compression and zero padding on the split image so that the target of the image to be detected does not produce excessive deformation and ensures that small targets are not missed due to image compression.

[0156] During image processing, the image to be detected is first segmented, and then the object detection model is run on each independent segmented block to identify and extract the object detection box, corresponding category label, and confidence score within each block. In order to ensure that these detection boxes can be accurately reflected in the global environment of the original image, the local coordinates within the segmented block are converted to the global coordinates of the original image. This conversion process depends on the exact position of the segmented block in the original image and the coordinate information of the detection box inside it.

[0157] After the conversion is completed, you may encounter multiple detection boxes that overlap or are close to each other. Algorithms such as non-maximum suppression (NMS) are used to merge these overlapping or close boxes. The NMS algorithm compares the confidence scores of adjacent boxes and retains the box with the highest score, while suppressing or deleting other boxes with lower scores, thereby effectively reducing duplicate detections.

[0158] Finally, by merging the detection boxes that represent the same target but are scattered in different segments, a more accurate and unified detection box is formed, which accurately covers the target area in the original image.

[0159] 706. Input at least one image to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0160] 707. Alert the target staff member based on the detection data.

[0161] In this embodiment, steps S701 to S704 are implemented in a similar manner to steps S101 to S104 in the above embodiment, and are not described in detail here. Steps S706 to S707 are similar to steps S105 to S106 in the above embodiment, and are not described in detail here.

[0162] See also Figure 8 The present application provides an embodiment of a detection device for fire retardant clothing, comprising:

[0163] A collection unit 801 is used to collect behavior videos of workers at at least two fire protection clothing detection points, each of which includes at least one collection device;

[0164] An acquisition unit 802 is used to acquire at least one to-be-detected image of the same target worker from the behavior video;

[0165] Optionally, the acquisition unit 802 is further configured to: uniformly collect at least two initialization images from the behavior video, and select a first image set of the same target worker from the at least two initialization images;

[0166] The target staff area of ​​each image in the first image set is detected, and images with unqualified area are screened out to generate at least two images to be collected.

[0167] Optionally, the method further includes a positioning unit 803: positioning a human posture frame of the target worker on the image to be detected;

[0168] Determine a target human body frame with the highest posture similarity in a human body frame library through the human body posture frame, wherein the target human body frame has a human body structure area;

[0169] Scaling the target human body frame according to the human body posture frame;

[0170] The fireproof clothing feature detection area is determined on the image to be detected through the human body structure area on the target human body frame, and the fireproof clothing feature detection area includes a helmet detection area, a top detection area, a bottom detection area, a boots detection area, a glove detection area and a belt detection area.

[0171] The grayscale mean of the image to be detected in the target background area is calculated, and the grayscale mean is determined as brightness data.

[0172] A determination unit 804 is used to determine the brightness data of the image to be detected;

[0173] Optionally, the determination unit 804 is further configured to: determine an initial background area of ​​the image to be detected through the fire protection clothing feature detection area;

[0174] Screening out the staff area other than the target staff in the initial background area to generate a target background area;

[0175] A processing unit 805 is used to perform enhancement processing on the image to be detected when the brightness data does not meet the preset brightness condition;

[0176] Optionally, the processing unit 805 is further used to: when the brightness data does not meet the preset brightness condition, use the CSNorm model to enhance the fire protection clothing feature detection area of ​​the image to be detected.

[0177] Optionally, the method further includes a segmentation unit 806: segmenting the image to be detected, and filling the segmented image so that the size of the image to be detected meets the requirements of the convolutional neural network model.

[0178] The output unit 807 is used to input at least one of the images to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and output at least one detection data;

[0179] Optionally, the output unit 807 is further used to: input at least one of the to-be-detected images into a pre-trained convolutional neural network model;

[0180] The convolutional neural network model is used to identify the reflective shape features of each fire retardant clothing feature detection area in at least one of the images to be detected, and at least one detection data is generated, wherein the detection data includes the recognition results of specific reflective shape features in each fire retardant clothing feature detection area of ​​one of the images to be detected.

[0181] The alarm unit 808 is used to alarm the target staff according to the detection data.

[0182] See also Fig. 9 , the present application also provides a detection device, comprising:

[0183] Processor 901, memory 902, input and output unit 903, bus 904;

[0184] The processor 901 is connected to the memory 902, the input and output unit 903 and the bus 904;

[0185] The memory 902 stores a program, and the processor 901 calls the program to execute any of the above detection methods.

[0186] The present application also relates to a computer-readable storage medium on which a program is stored, wherein when the program is run on a computer, the computer is caused to execute any of the above detection methods.

[0187] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0188] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0189] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0190] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0191] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. A method for detecting fire retardant clothing, characterized in that: The detection method comprises: Collecting behavior videos of workers at at least two fire-resistant clothing inspection points, each of which includes at least one collection device; Acquire at least one to-be-detected image of the same target worker from the behavior video; Determining brightness data of the image to be detected; When the brightness data does not meet the preset brightness condition, enhancing the image to be detected; Inputting at least one of the images to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and outputting at least one detection data; The target staff is alerted based on the detection data.

2. The detection method according to claim 1, characterized in that: After obtaining at least one image to be detected of the same target worker from the behavior video and before determining the brightness data of the image to be detected, the detection method further includes: Select image data of the same scene collected by the same camera device, use the frame difference method to process the two frames of images, and judge whether there is a target in the image through the foreground image area retained after the frame difference method. And set the threshold according to the shape and area of ​​the foreground target area to perform background interference filtering. The fireproof clothing feature detection area is determined on the image to be detected through the reserved foreground area, and the fireproof clothing feature detection area includes a fireproof helmet detection area, a fireproof clothing detection area, a fireproof pants detection area, a boots detection area, a glove detection area and a belt detection area.

3. The detection method according to claim 2, characterized in that: Determining brightness data of the image to be detected includes: Determining the initial background area of ​​the image to be detected through the fireproof clothing feature detection area; Screening out the staff area other than the target staff in the initial background area to generate a target background area; The grayscale mean of the image to be detected in the target background area is calculated, and the grayscale mean is determined as brightness data.

4. The detection method according to claim 3, characterized in that: Each component of the fireproof clothing is provided with a reflective area, and the reflective shape characteristics of the reflective area of ​​each component are different; Inputting at least one of the images to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and outputting at least one detection data, including: Inputting at least one of the images to be detected into a pre-trained convolutional neural network model; The convolutional neural network model is used to identify the reflective shape features of each fire retardant clothing feature detection area in at least one of the images to be detected, and at least one detection data is generated, wherein the detection data includes the recognition results of specific reflective shape features in each fire retardant clothing feature detection area of ​​one of the images to be detected.

5. The detection method according to any one of claims 1 to 4, characterized in that: The step of obtaining at least one image to be detected of the same target worker from the behavior video includes: uniformly collecting at least two initialization images from the behavior video, and selecting foreground area images of the same target worker as a first image set from the at least two initialization images by using a frame difference method; The target staff area of ​​each image in the first image set is detected, and images with unqualified area are screened out to generate at least two images to be collected.

6. The detection method according to claim 2, characterized in that: When the brightness data does not meet the preset brightness condition, the image to be detected is enhanced, including: When the brightness data does not meet the preset brightness condition, the CSNorm model is used to enhance the fireproof clothing feature detection area of ​​the image to be detected.

7. The detection method according to claim 1, characterized in that: When the brightness data does not meet the preset brightness condition, after the image to be detected is enhanced, before at least one image to be detected is input into a pre-trained convolutional neural network model for fire protection clothing feature detection and before at least one detection data is output, the detection method includes: The image to be detected is randomly cropped and segmented, and the segmented image is geometrically compressed and filled with zeros to prevent the target of the image to be detected from being deformed too much and to ensure that small targets are not missed due to image compression.

8. A detection device for fire retardant clothing, characterized in that: The detection device comprises: A collection unit, used to collect behavior videos of workers at at least two fire protection clothing detection points, each of which includes at least one collection device; An acquisition unit, used for acquiring at least one to-be-detected image of the same target worker from the behavior video; A determination unit, used to determine the brightness data of the image to be detected; A processing unit, configured to perform enhancement processing on the image to be detected when the brightness data does not meet a preset brightness condition; An output unit, used for inputting at least one of the images to be detected into a pre-trained convolutional neural network model to detect the characteristics of the fireproof clothing, and outputting at least one detection data; An alarm unit is used to alarm the target staff according to the detection data.

9. A detection device, characterized in that: The device comprises: Processor, memory, input-output unit, and bus; The processor is connected to the memory, the input and output unit, and the bus; The memory stores a program, and the processor calls the program to execute the detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a program stored thereon, wherein the program, when executed on a computer, performs the detection method according to any one of claims 1 to 7.

Images