CN114998830A - Wearing detection method and system for safety helmet of transformer substation personnel - Google Patents
Wearing detection method and system for safety helmet of transformer substation personnel Download PDFInfo
- Publication number
- CN114998830A CN114998830A CN202210551125.8A CN202210551125A CN114998830A CN 114998830 A CN114998830 A CN 114998830A CN 202210551125 A CN202210551125 A CN 202210551125A CN 114998830 A CN114998830 A CN 114998830A
- Authority
- CN
- China
- Prior art keywords
- monitoring image
- detection result
- safety helmet
- model
- preprocessed
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/041—Abduction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention belongs to the field of image identification and detection, and provides a method and a system for detecting wearing of safety helmets of transformer substation personnel, which comprises the steps of obtaining an original monitoring image acquired by video monitoring equipment and preprocessing the original monitoring image; based on the preprocessed monitoring image and the original monitoring image, detecting by using a pre-trained safety helmet wearing detection model to obtain a preprocessed monitoring image detection result and an original monitoring image detection result; fusing the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result; according to the improved YOLOv5 series model, a new attention module CAM (computer-aided manufacturing) attention module is added on the basis of the existing YOLOv5 model, so that the position information in the feature space can be integrated into the channel attention, the model can sense a larger area, and the accuracy of the model on small target and occlusion recognition is effectively improved.
Description
Technical Field
The invention belongs to the technical field of image identification and detection, and particularly relates to a method and a system for detecting wearing of safety helmets of personnel in a transformer substation.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Each industry puts high requirements on the safe operation of constructors. The degree of injury to constructors caused by dangerous accidents can be reduced by wearing the safety helmet, and the threat to life and property is effectively reduced. However, it is difficult to ensure that the worker wears the helmet all the time by using a monitoring system or checking the wearing condition of the helmet of the worker on site through human eyes, and it takes a lot of time and labor cost, and it is difficult to achieve the expected effect in actual implementation, which does not meet the requirement of modern construction safety management.
The transformer substation monitoring is an important means for ensuring the safe operation of the transformer substation, and the monitoring of whether a construction inspection worker wears a safety helmet is a key component of an intelligent monitoring system of the transformer substation.
Conventional target detection techniques typically employ manual selection of features and design and training of classifiers based on a particular detection object. The method has the advantages of strong subjectivity, complex design process, poor universality and great limitation in engineering application.
In recent years, a target detection method based on deep learning is widely applied to an intelligent monitoring system. Two target detection modes based on deep learning are mainly two, two-stage and one-stage. The former is an R-CNN series target detection algorithm based on a candidate region, and the candidate region needs to be generated firstly, and then the classification and regression operation are carried out on the candidate region; the latter are the YOLO and SSD algorithms, which use only one CNN network to directly predict the classes and locations of different targets. Compared with the two-stage series algorithm, the one-stage series algorithm can realize real-time detection, but has lower precision.
Disclosure of Invention
In order to solve the problems, the invention provides a transformer substation personnel safety helmet wearing detection method and system, and the improved YOLOv5 series model is additionally provided with a new attention module CAM (computer-aided manufacturing) attention module on the basis of the existing YOLOv5 model, so that the position information in the feature space can be integrated into the channel attention, the model can sense a larger area, and the accuracy of the model on small target and shielding identification is effectively improved.
According to some embodiments, a first aspect of the present invention provides a method for detecting wearing of a safety helmet of a transformer substation person, which adopts the following technical solutions:
a transformer substation personnel safety helmet wearing detection method comprises the following steps:
acquiring an original monitoring image acquired by video monitoring equipment and preprocessing the original monitoring image;
based on the preprocessed monitoring image and the original monitoring image, detecting by using a pre-trained safety helmet wearing detection model to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
fusing the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result;
the helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase a perception area of the YOLOv5 network model.
Further, the acquiring and preprocessing of the original monitoring image acquired by the video monitoring device includes:
acquiring an original monitoring image acquired by video monitoring equipment;
and segmenting the original monitoring image into blocks containing the overlapping area to obtain the preprocessed monitoring image.
Further, the attention module performs an image processing process, including:
encoding the input feature vectors in a width and height one-way dimension by average pooling to obtain the input feature vectors in the width direction and the input feature vectors in the height direction;
fusing the input characteristic vector in the width direction and the input characteristic vector in the height direction;
performing convolution and dimension reduction on the fused feature vectors to obtain a feature map containing space related information in the vertical direction and the horizontal direction;
converting a feature map containing space related information in the vertical direction and the horizontal direction into two independent feature vectors;
transforming the two independent feature vectors by using two depth separable convolutions respectively to obtain two transformed independent feature vectors and obtain attention weights of the two independent feature vectors;
and applying the attention weights of the two independent feature vectors to the input feature vector to obtain an output feature vector of the whole attention module.
Further, applying attention weights of the two independent feature vectors to the input feature vector to obtain an output feature vector of the whole attention module, specifically:
wherein, g h And g w Are two independent feature vectors, x c (i, j) is the input feature vector, y c (i, j) is the output feature vector; i, j are characteristicsCoordinate index of the vector, c denotes channel index.
Further, a process of training a headgear wearing detection model includes:
acquiring an original monitoring image acquired by video monitoring equipment;
labeling an original monitoring image by using labelme to obtain a labeled monitoring image;
segmenting the original monitoring image and the marked monitoring image to obtain a preprocessed monitoring image and a preprocessed marked monitoring image;
zooming the preprocessed monitoring image;
performing data enhancement on the zoomed monitoring image by adopting Mosaic, random zooming, random cutting and perspective transformation;
inputting the monitoring image subjected to data enhancement and the preprocessed marked monitoring image into a safety helmet wearing detection model;
and updating the parameters of the safety helmet wearing detection model according to the oscillation condition of the historical gradient and the real historical gradient after filtering oscillation, wherein the loss function gradually tends to 0, and the model converges to obtain the trained safety helmet wearing detection model.
Further, for the labeled monitoring image, a Kmeans + + clustering algorithm is adopted to obtain 9 anchor frames (anchors), and the anchor frames are sequentially arranged from small to large and are uniformly distributed on feature maps with 3 scales.
Further, the pre-processed monitoring image detection result and the original monitoring image detection result are fused to generate a final detection result, which specifically comprises:
fusing the detection result of the preprocessed monitoring image with the detection result of the monitoring image to generate a final detection result;
and if the intersection between the target frames of the two frames is larger than the set threshold, deleting the detection result with lower score.
According to some embodiments, a second aspect of the present invention provides a transformer substation personnel safety helmet wearing detection system, which adopts the following technical solutions:
a transformer substation personnel safety helmet wearing detection system comprises:
the image acquisition module is configured to acquire an original monitoring image acquired by the video monitoring equipment and perform preprocessing;
the image detection module is configured to detect by using a pre-trained safety helmet wearing detection model based on the preprocessed monitoring image and the original monitoring image to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
the detection result fusion module is configured to fuse the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result;
the helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase a perception area of the YOLOv5 network model.
According to some embodiments, a third aspect of the invention provides a computer-readable storage medium.
A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, carries out the steps in a substation personnel headgear wear detection as described in the first aspect above.
According to some embodiments, a fourth aspect of the invention provides a computer apparatus.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of a method for detecting wearing of personal safety helmets of substations as described in the first aspect above when executing the program.
Compared with the prior art, the invention has the beneficial effects that:
according to the improved YOLOv5 series model, a new attention module CAM (computer-aided manufacturing) attention module is added on the basis of the existing YOLOv5 model, so that the position information in the feature space can be integrated into the channel attention, the model can sense a larger area, and the accuracy of the model on small target and occlusion recognition is effectively improved.
The image preprocessing method can be widely applied to lightweight model training and reasoning frameworks, and effectively improves the accuracy of the model for detecting the small target.
The improved YOLOv5 model can effectively improve the identification accuracy of the model to small targets and occlusion conditions. By combining the image preprocessing method, the missing detection rate of the actual requirements for wearing detection of the safety helmet of the transformer substation is 3.47%, the false detection rate is 5.38%, and the accuracy is 97.21%, so that the actual requirements for wearing detection of the safety helmet of the transformer substation are met.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a flowchart of a method for detecting wearing of a safety helmet of a transformer substation personnel according to an embodiment of the present invention;
FIG. 2 is a flow chart of a pre-process according to an embodiment of the present invention;
FIG. 3 is a fused graph of the test results according to the embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating fusion of detection results according to an embodiment of the present invention;
FIG. 5 is a diagram of an attention module network architecture according to an embodiment of the present invention;
fig. 6 is a structural diagram of a CAM module added to a backbone network in the YOLOv5 model according to an embodiment of the present invention;
FIG. 7 is a block diagram of a neck networking CAM bank in the YOLOv5 model according to an embodiment of the present invention;
fig. 8 is a structural diagram of a CAM module added to a header network in the YOLOv5 model according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example one
As shown in fig. 1, the present embodiment provides a method for detecting wear of a safety helmet of a transformer substation, and the present embodiment is exemplified by applying the method to a server, and it may be understood that the method may also be applied to a terminal, and may also be applied to a system and a terminal, and is implemented through interaction between the terminal and the server. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a network server, cloud communication, middleware service, a domain name service, a security service CDN, a big data and artificial intelligence platform, and the like. The terminal may be, but is not limited to, a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and the like. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein. In this embodiment, the method includes the steps of:
acquiring an original monitoring image acquired by video monitoring equipment and preprocessing the original monitoring image;
based on the preprocessed monitoring image and the original monitoring image, detecting by using a pre-trained safety helmet wearing detection model to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
fusing the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result;
the helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase a perception area of the YOLOv5 network model.
Specifically, the method of this embodiment includes the following steps:
1. the scene image of a person wearing the safety helmet to be detected is derived from video monitoring equipment erected in a transformer substation, video information acquired by the monitoring equipment in real time is acquired and sent to an analysis server, and a video stream acquired is intercepted to obtain a monitoring image, so that real-time acquisition and accumulation of data are realized;
2. establishing a transformer substation personnel safety helmet wearing label data set through the open source data set and data accumulated by the monitoring equipment, wherein the label data set is used for improving the training of a YOLOv5 transformer substation safety helmet wearing detection model;
3. monitoring image I acquired by capturing and intercepting video monitoring equipment i Performing a preprocessing operation to obtain a preprocessed image I' i,j 。
4. Pre-processed image I' i,j Inputting a safety helmet wearing detection model to obtain a preprocessed image I' i,j Corresponding detection result R i,j And fusing the detection results to generate a final detection result.
Data set preparation
The preparation of the data set comprises two parts, and labelme is used for marking the data collected and accumulated by the monitoring equipment. Specifically, the substation personnel head area with the safety helmet is labeled "Wear _ helmet", and the substation personnel head area without the safety helmet is labeled "No _ helmet"; and mapping the annotation file types for the open source data set, and revising the open source data with inconsistent annotation modes.
Image pre-processing
Due to the fact that the image resolution is larger and larger (1920x1080,2K, 4K and 8K of unmanned aerial vehicles and the like), the difference between the original image resolution and the model input resolution is larger and larger. And the resolution of the input model is limited by computational power, so that the problem of inevitable detail loss is caused. In order to further improve the recognition accuracy of the model for the small and medium targets in the high-resolution camera, image preprocessing operation is required.
As shown in fig. 2, the preprocessing operations may act simultaneously in the training and reasoning phases.
In the inference stage, segmenting an original resolution image into small blocks containing an overlapping region, inputting the segmented image and the whole frame into a safety helmet wearing detection model to obtain detection results of the segmented image and the whole frame, and fusing the results;
the training stage is different from the reasoning stage in that the image cutting affects the labeled file, and the labeled file of the original image is processed to keep consistent with the cut image.
Meanwhile, the boundary box falling on the clipping boundary target is considered. Through the preprocessing operation, the size of the target of the cut image is increased relative to the original image when the model is input, so that the loss of details of small and medium targets caused by image scaling is avoided, and the missing detection of the small targets by the model is reduced. It should be noted that, since the original image is also subjected to model training and reasoning, a large target can be detected for the detected image.
Since the image is subjected to the blocking operation, that is, one image becomes a plurality of images. For example, an 800 × 600 image is divided into 2 × 2 images, the size of each image should be about 400 × 300 (overlapping is not considered), and if there is an object of [500,330], [550,400] in the annotation file of the original image, it will fall outside the image after the block division, so the annotation file also needs to be modified to generate 4 annotation files. That is, in the above example, the original image and the markup file, and the 4 images and 4 markup files after being partitioned, which are 5 images and 5 markup files, are included in the training stage.
Notably, clipping the number of patches to the original image linearly increases the complexity of the overall detection framework. Therefore, the method is suitable for lightweight networks to meet real-time reasoning without increasing memory usage. Based on the preprocessing mode, the detection precision of the small target can be obviously improved. The block grid is typically chosen to be 3x 2.
Safety helmet wearing detection model establishment
In order to further reduce the time-consuming influence caused by the image preprocessing operation and simultaneously improve the capability of the model for detecting the small target. The scheme is preferably but not limited to a modified YOLOv5 series model, and a new Attention module cam (coordinate Attention module) is added on the basis of the existing YOLOv5 model, so that the position information in the feature space can be integrated into the channel Attention, and the model can sense a larger area. The attention module network structure is shown in fig. 5. Specifically, first, the input feature vector X having dimensions H × W × C is encoded in one dimension in the width and height directions using average pooling with pooling sizes of (H,1) and (1, W). The output of each channel with the height h is shown as the formula (1); similarly, the output of each channel with width w is shown in formula (2).
Wherein the content of the first and second substances,
represents the output of channel c at height h, W represents the width of the input feature vector X;
the output of channel c at width w is indicated and H represents the height of the input feature vector X.
For the two direction perception features z obtained above h And z w And (3) performing CONCAT operation, then performing dimensionality reduction by using 1x1 convolution to obtain a feature map f containing vertical and horizontal direction space correlation information, and outputting the feature map f as shown in the formula (3).
f＝δ(F1([z h ,z w ])) (3)
Wherein F1 represents a 1x1 convolution operation; δ represents the nonlinear activation function.
Then, f is transformed into two independent feature vectors f h And f w And transforming the feature vectors by using two depth separable convolutions (Dpthwise separable Cnvolution), respectively, to obtain transformed feature vectors g h And g w Has the same channel number as the input feature vector X, as shown in formula (4) and formula (5), and obtains g h And g w Is the attention weight.
g h ＝σ(F h (f h )) (4)
g w ＝σ(F w (f w )) (5)
Wherein, F h And F w Respectively representing depth separable convolutions applied to two independent feature vectors; σ denotes Sigmoid activation function.
Therefore, attention weight is applied to the input feature vector X to obtain the output of the entire attention module, as shown in equation (6).
Wherein x is c (i, j) is the input feature vector, y c (i, j) is the output feature vector; i, j is the coordinate index of the feature vector, c represents the channel index; that is, i is the index over the width of the feature vector and j is the index over the height of the feature vector.
The attention module can be integrated into a plurality of positions such as a trunk structure (backbone), a neck (neck) and a head (head) of a YOLOv5 model, taking a YOLOv5s model structure as an example, a structure for adding CAM in the trunk structure, the neck and 3 predicted regions of a model network structure is shown in fig. 6, 7 and 8.
Specifically, the following:
a backbone network composed of a Focus structure, a CBL structure, a CSP structure, and an SPP structure, and to which an attention module is connected after each CSP1_3 structure;
the neck network is composed of a CBL structure, a CSP structure and an up-sampling structure, and an attention module is added after each feature fusion from bottom to top;
the head network is formed by decoupling convolution module conv and YOLO, and attention modules are connected to the features after the features with different scales are fused.
As shown in fig. 6, the CAM module is integrated at CSP1_3 (feature fusion) of the YOLOV5s backbone; as shown in fig. 7, the CAM module is integrated behind the CONCAT layer at the neck of YOLOV5 s; as shown in FIG. 8, the YOLOv5s header (head) is respectively integrated with CAM for different scale features and then input into the yolo-head structure.
Further, based on the helmet wearing data set, 9 anchor frames (anchors) are obtained by adopting a Kmeans + + clustering algorithm and are sequentially arranged from small to large and uniformly distributed on a characteristic diagram with 3 scales.
The use of CIoU _ Loss for bounding box Loss better measures the case where the predicted box intersects the target box. As shown in equation (7).
Wherein the content of the first and second substances,
a and B are respectively a prediction frame and a target frame; ρ represents the Euclidean distance between two boxes, b and b gt The center points of the two rectangular boxes are indicated, and c represents the distance of the diagonal of the closure area of the two rectangular boxes.
And the CIoU-NMS is used for replacing NMS, so that the identification degree of the shielding target is improved.
Helmet wearing detection model training process, specifically:
as the preprocessing operation described above, the image and its annotation file are processed in blocks;
zooming the image to 640x640, wherein the original image scale is kept in the zooming process;
adopting data enhancement modes such as Mosaic, random zooming, random cutting, perspective transformation and the like; in the training process, updating model parameters by using an Adam optimizer according to the oscillation condition of the historical gradient and the real historical gradient after filtering oscillation, gradually leading the loss function to be close to 0, converging the model, and obtaining a safety helmet wearing detection model;
result fusion
As shown in fig. 3 and 4, in the inference process, since the image is subjected to the blocking process, it is necessary to fuse the detection results of the different block images and the detection result of the original image to generate the final detection result. Specifically, due to the overlap between the different tile images and the original image, a repeatedly detected object may appear in the initial result. For the target with repeated detection in the result, if the intersection between the target frames is more than 25%, the result with higher score is accepted, and another result is deleted from the detection list. It should be noted that in the merging process, the result score of the small object in the block image is higher than that of the original image, and the result score of the large object with the size equivalent to that of the block image is opposite to that of the original image. This is also why it is necessary to reason about both the tile image and the original image.
Example two
This embodiment provides a detecting system is worn to personnel's safety helmet of transformer substation, includes:
the image acquisition module is configured to acquire an original monitoring image acquired by the video monitoring equipment and perform preprocessing;
the image detection module is configured to detect by using a pre-trained safety helmet wearing detection model based on the preprocessed monitoring image and the original monitoring image to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
the detection result fusion module is configured to fuse the detection result of the preprocessed monitoring image with the detection result of the original monitoring image to generate a final detection result;
the helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase a perception area of the YOLOv5 network model.
The modules are the same as the corresponding steps in the implementation example and application scenarios, but are not limited to the disclosure of the first embodiment. It should be noted that the modules described above as part of a system may be implemented in a computer system such as a set of computer-executable instructions.
In the foregoing embodiments, the descriptions of the embodiments have different emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
The proposed system can be implemented in other ways. For example, the above-described system embodiments are merely illustrative, and for example, the division of the above-described modules is merely a logical division, and in actual implementation, there may be other divisions, for example, multiple modules may be combined or integrated into another system, or some features may be omitted, or not executed.
EXAMPLE III
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in a method for detecting wearing of a personal safety helmet of a substation as described in the first embodiment above.
Example four
The embodiment provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, the steps in the method for detecting the wearing of the safety helmet of the transformer substation personnel are implemented as in the first embodiment.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A transformer substation personnel safety helmet wearing detection method is characterized by comprising the following steps:
acquiring an original monitoring image acquired by video monitoring equipment and preprocessing the original monitoring image;
based on the preprocessed monitoring image and the original monitoring image, detecting by using a pre-trained safety helmet wearing detection model to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
fusing the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result;
the safety helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase the perception area of the YOLOv5 network model.
2. The method for detecting wearing of the safety helmet of the substation personnel as claimed in claim 1, wherein the obtaining and preprocessing of the original monitoring image acquired by the video monitoring device comprises:
acquiring an original monitoring image acquired by video monitoring equipment;
and segmenting the original monitoring image into blocks containing the overlapping area to obtain the preprocessed monitoring image.
3. The method for detecting wearing of safety helmets of transformer stations according to claim 1, wherein the attention module performs an image processing process comprising:
encoding the input feature vectors in a width and height one-way dimension by average pooling to obtain the input feature vectors in the width direction and the input feature vectors in the height direction;
fusing the input characteristic vector in the width direction and the input characteristic vector in the height direction;
performing convolution dimensionality reduction on the fused input feature vectors to obtain a feature map containing space related information in the vertical direction and the horizontal direction;
converting a feature map containing space related information in the vertical direction and the horizontal direction into two independent feature vectors;
transforming the two independent feature vectors by using two depth separable convolutions respectively to obtain two transformed independent feature vectors and obtain attention weights of the two independent feature vectors;
and applying the attention weights of the two independent feature vectors to the input feature vector to obtain an output vector of the whole attention module.
4. The transformer substation personnel safety helmet wearing detection method according to claim 3,
applying the attention weights of the two independent feature vectors to the input feature vector to obtain an output vector of the whole attention module, specifically comprising the following steps:
wherein, g h And g w Are two independent feature vectors, x c (i, j) is the input feature vector, y c (i, j) is the output feature vector; i, j are the coordinate indices of the feature vectors, and c denotes the channel index.
5. The transformer substation personnel safety helmet wearing detection method according to claim 1, wherein the process of training the safety helmet wearing detection model comprises the following steps:
acquiring an original monitoring image acquired by video monitoring equipment;
labeling the original monitoring image by using label to obtain a labeled monitoring image;
segmenting the original monitoring image and the marked monitoring image to obtain a preprocessed monitoring image and a preprocessed marked monitoring image;
zooming the preprocessed monitoring image;
performing data enhancement on the zoomed monitoring image by adopting Mosaic, random zooming, random cutting and perspective transformation;
inputting the monitoring image subjected to data enhancement and the preprocessed marked monitoring image into a safety helmet wearing detection model;
and updating the parameters of the safety helmet wearing detection model according to the oscillation condition of the historical gradient and the real historical gradient after filtering oscillation, wherein the loss function gradually tends to 0, and the model converges to obtain the trained safety helmet wearing detection model.
6. The transformer substation personnel safety helmet wearing detection method according to claim 5, characterized in that for the marked monitoring image, 9 anchor frames (anchors) are obtained by adopting a Kmeans + + clustering algorithm and are arranged from small to large in sequence and are uniformly distributed on a characteristic diagram with 3 scales.
7. The method for detecting wearing of the safety helmet of the substation personnel according to claim 1, wherein the preprocessed monitoring image detection result and the original monitoring image detection result are fused to generate a final detection result, and specifically the method comprises the following steps:
fusing the detection result of the preprocessed monitoring image with the detection result of the original monitoring image to generate a final detection result;
and if the intersection between the target frames of the two frames is larger than the set threshold, deleting the detection result with low score.
8. The utility model provides a detecting system is worn to personnel's safety helmet of transformer substation which characterized in that includes:
the image acquisition module is configured to acquire an original monitoring image acquired by the video monitoring equipment and perform preprocessing;
the image detection module is configured to detect by using a pre-trained safety helmet wearing detection model based on the preprocessed monitoring image and the original monitoring image to obtain a preprocessed monitoring image detection result and an original monitoring image detection result;
the detection result fusion module is configured to fuse the detection result of the preprocessed monitoring image and the detection result of the original monitoring image to generate a final detection result;
the safety helmet detection model adopts a YOLOv5 network model, a trunk network, a neck network and a head network in the YOLOv5 network model are respectively integrated with a plurality of attention modules, and the attention modules integrate position information in a feature space into channel attention and increase the perception area of the YOLOv5 network model.
9. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the steps of a method for detecting the wearing of a safety helmet of a substation personnel according to any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps in a method for substation personal safety helmet wear detection as claimed in any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210551125.8A CN114998830A (en) | 2022-05-20 | 2022-05-20 | Wearing detection method and system for safety helmet of transformer substation personnel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210551125.8A CN114998830A (en) | 2022-05-20 | 2022-05-20 | Wearing detection method and system for safety helmet of transformer substation personnel |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114998830A true CN114998830A (en) | 2022-09-02 |
Family
ID=83027346
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210551125.8A Pending CN114998830A (en) | 2022-05-20 | 2022-05-20 | Wearing detection method and system for safety helmet of transformer substation personnel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114998830A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630350A (en) * | 2023-07-26 | 2023-08-22 | 瑞茜时尚(深圳)有限公司 | Wig wearing monitoring management method and system |
| CN116958907A (en) * | 2023-09-18 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Method and system for inspecting surrounding hidden danger targets of gas pipeline |
| CN116958883A (en) * | 2023-09-15 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Safety helmet detection method, system, storage medium and electronic equipment |
| CN117036327A (en) * | 2023-08-22 | 2023-11-10 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Protective article inspection method, system, equipment and medium |
-
2022
- 2022-05-20 CN CN202210551125.8A patent/CN114998830A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630350A (en) * | 2023-07-26 | 2023-08-22 | 瑞茜时尚(深圳)有限公司 | Wig wearing monitoring management method and system |
| CN116630350B (en) * | 2023-07-26 | 2023-10-03 | 瑞茜时尚(深圳)有限公司 | Wig wearing monitoring management method and system |
| CN117036327A (en) * | 2023-08-22 | 2023-11-10 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Protective article inspection method, system, equipment and medium |
| CN117036327B (en) * | 2023-08-22 | 2024-03-12 | 广州市疾病预防控制中心(广州市卫生检验中心、广州市食品安全风险监测与评估中心、广州医科大学公共卫生研究院) | Protective article inspection method, system, equipment and medium |
| CN116958883A (en) * | 2023-09-15 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Safety helmet detection method, system, storage medium and electronic equipment |
| CN116958883B (en) * | 2023-09-15 | 2023-12-29 | 四川泓宝润业工程技术有限公司 | Safety helmet detection method, system, storage medium and electronic equipment |
| CN116958907A (en) * | 2023-09-18 | 2023-10-27 | 四川泓宝润业工程技术有限公司 | Method and system for inspecting surrounding hidden danger targets of gas pipeline |
| CN116958907B (en) * | 2023-09-18 | 2023-12-26 | 四川泓宝润业工程技术有限公司 | Method and system for inspecting surrounding hidden danger targets of gas pipeline |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114998830A (en) | Wearing detection method and system for safety helmet of transformer substation personnel | |
| CN109241985B (en) | Image identification method and device | |
| CN112200043A (en) | Intelligent danger source identification system and method for outdoor construction site | |
| CN111222478A (en) | Construction site safety protection detection method and system | |
| CN109886159B (en) | Face detection method under non-limited condition | |
| CN104504377A (en) | Bus passenger crowding degree identification system and method | |
| CN113762201A (en) | Mask detection method based on yolov4 | |
| CN111460988A (en) | Illegal behavior identification method and device | |
| Jin et al. | DWCA-YOLOv5: An improve single shot detector for safety helmet detection | |
| US20230017578A1 (en) | Image processing and model training methods, electronic device, and storage medium | |
| CN111539320B (en) | Multi-view gait recognition method and system based on mutual learning network strategy | |
| Zhao et al. | Detection and location of safety protective wear in power substation operation using wear-enhanced YOLOv3 algorithm | |
| CN112668672A (en) | TensorRT-based target detection model acceleration method and device | |
| CN115965578A (en) | Binocular stereo matching detection method and device based on channel attention mechanism | |
| CN113326773A (en) | Recognition model training method, recognition method, device, equipment and storage medium | |
| CN115620208A (en) | Power grid safety early warning method and device, computer equipment and storage medium | |
| Yang et al. | Safety helmet wearing detection based on an improved YOLOv3 scheme | |
| CN113936299A (en) | Method for detecting dangerous area in construction site | |
| CN116597411A (en) | Method and system for identifying traffic sign by unmanned vehicle in extreme weather | |
| CN115601674A (en) | Power transmission project smoke hidden danger identification method and device and storage medium | |
| CN115631457A (en) | Man-machine cooperation abnormity detection method and system in building construction monitoring video | |
| CN115100592A (en) | Method and device for identifying hidden danger of external damage of power transmission channel and storage medium | |
| Yu et al. | An Algorithm for Target Detection of Engineering Vehicles Based on Improved CenterNet. | |
| Daniya et al. | ICSA-ECNN based image forgery detection in face images | |
| Yu et al. | Safety Helmet Wearing Detection Based on Super-resolution Reconstruction |
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 |