CN110889376A - Safety helmet wearing detection system and method based on deep learning - Google Patents
Safety helmet wearing detection system and method based on deep learning Download PDFInfo
- Publication number
- CN110889376A CN110889376A CN201911190920.3A CN201911190920A CN110889376A CN 110889376 A CN110889376 A CN 110889376A CN 201911190920 A CN201911190920 A CN 201911190920A CN 110889376 A CN110889376 A CN 110889376A
- Authority
- CN
- China
- Prior art keywords
- human body
- safety helmet
- detection
- image
- detection result
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Helmets And Other Head Coverings (AREA)
Abstract
The invention relates to the technical field of image recognition, in particular to a safety helmet wearing detection system and a method based on deep learning, wherein the system comprises the following components: the system comprises a data acquisition module, a human body detection result analysis module, an image segmentation module, a safety helmet detection module, an early warning module and a model training module; the method comprises the following steps: step S1, the safety helmet wearing detection system acquires a monitoring image at the monitoring equipment; step S2, the safety helmet wearing detection system carries out human body target detection on the monitored image and obtains a human body target detection result; and step S3, the safety helmet wearing detection system carries out safety helmet identification detection on each human body image in the human body target detection result to judge whether each human body image has a safety helmet or not and obtain the safety helmet detection result.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a safety helmet wearing detection system and method based on deep learning.
Background
In a construction site, a workshop, a dock and the like, workers need to wear safety helmets to prevent articles from falling off to cause injuries to the workers, but often the workers do not wear the safety helmets for various reasons, and in order to improve production safety and protect life safety of the workers, an effective monitoring and management method is needed to supervise the workers wearing the safety helmets.
In the prior art, the following methods are generally adopted to automatically identify and detect whether a worker wears a safety helmet:
in the prior art, a safety helmet detection system directly detects workers, a safety helmet and a safety belt in an acquired monitoring picture by adopting a convolutional neural network, and then judges whether the safety helmet is correctly worn and whether the safety belt is fastened or not according to the mutual relation among the safety helmet, the safety belt and a human body. The method directly detects the safety helmet from the image, and the safety helmet is small and does not have abundant texture information, so that the detection accuracy of the safety helmet is low. In addition, the monitoring equipment is generally used for shooting downwards in a bent manner, the head of a shot human body is not obvious due to interference of external factors in some cases, or the head of the human body is completely shielded by the safety helmet, and at the moment, the convolutional neural network is adopted to judge whether the safety helmet is worn to the head by a worker obviously and cannot be achieved by calculating the overlapping rate of the head of the human body and the safety helmet.
Disclosure of Invention
In order to solve the technical problems, the invention provides a safety helmet wearing detection system based on deep learning, which integrates a human body target detection method and a safety helmet color information detection method, can effectively detect whether workers in a monitored image wear the safety helmet in a fitting manner, and can effectively reduce the omission of the safety helmet caused by human body occlusion in the monitored image or virtual detection between people in a crowd.
The technical problem solved by the invention can be realized by adopting the following technical scheme:
a deep learning based headgear wear detection system comprising:
the data acquisition module is used for acquiring a monitoring image at the monitoring equipment;
a human body detection module connected with the data acquisition module, wherein a human body target detection model and a human body skeleton key point detection model are preset in the human body detection module, and the human body detection module is used for detecting a human body target of the monitored image based on the human body target detection model and obtaining a human body target detection result;
the human body detection module is also used for detecting human body bone key points of a human body image existing in the human body target detection result based on the human body bone key point detection model and obtaining a human body bone key point detection result;
the human body detection result analysis module is connected with the human body detection module and is used for combining the human body target detection result and the human body bone key point detection result and obtaining a human body detection result through comprehensive analysis;
the image segmentation module is connected with the human body detection result analysis module and is used for segmenting each human body image in the human body detection result into a plurality of human body region pictures related to each human body image and storing the human body region pictures;
the safety helmet detection module is connected with the image segmentation module and used for identifying and detecting whether a safety helmet exists in each human body region picture according to a preset safety helmet detection model and obtaining a safety helmet detection result;
and the early warning module is connected with the safety helmet detection module and used for carrying out corresponding prompt warning according to the safety helmet detection result.
Preferably, the safety helmet wearing detection system further comprises a model training module, which is respectively connected to the human body detection module and the safety helmet detection module, and is used for training to form the human body target detection model, the human body skeleton key point detection model and the safety helmet detection model, updating the human body target detection model and the human body skeleton key point detection model in the human body detection module, and updating the safety helmet detection model in the safety helmet detection module.
Preferably, the safety helmet detection module specifically includes:
the first safety helmet identification unit is used for carrying out safety helmet detection on each human body area picture based on the safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture or not;
the second safety helmet identification unit is connected with the first safety helmet identification unit and used for further calculating the color information of the image of the area where the suspected safety helmet is located in the human body area image to obtain a color information calculation result corresponding to the image of the area where the suspected safety helmet is located;
and the safety helmet detection result forming unit is connected with the first safety helmet identification unit and the second safety helmet identification unit and used for integrating the suspected safety helmet area identification result obtained by the first safety helmet identification unit and the color information calculation result corresponding to the suspected safety helmet area obtained by the second safety helmet identification unit to obtain a final detection result of the safety helmet as the safety helmet detection result and outputting the final detection result.
Preferably, the human body target detection is to perform coordinate position detection on coordinates of each human body image in the monitoring image, and the human body skeleton key point detection is to perform coordinate position detection on coordinates of each human body skeleton key point on each human body image on the corresponding human body image.
The invention provides a safety helmet wearing detection method based on deep learning, which is realized by a safety helmet wearing detection system and comprises the following steps:
step S1, the safety helmet wearing detection system obtains the monitoring image at the monitoring device;
step S2, the safety helmet wearing detection system carries out human body target detection on the monitoring image and obtains the human body target detection result;
step S3, the safety helmet wearing detection system performs safety helmet identification detection on each human body image in the human body target detection result to determine whether a safety helmet exists in each human body image, and obtains the safety helmet detection result.
Preferably, the step S2 specifically includes the following steps:
step S21, the safety helmet wearing detection system carries out human body target detection on the monitoring image based on a preset target detection model, and obtains the human body target detection result;
step S22, the safety helmet wearing detection system detects human skeleton key points of each human body image in the human body target detection result based on a preset human skeleton key point detection model, and obtains the human skeleton key point detection result;
and step S23, the safety helmet wearing detection system obtains the human body detection result through comprehensive analysis according to the human body target detection result and the human body bone key point detection result.
Preferably, the step S3 specifically includes the following steps:
step S31, the headgear wearing detection system divides each human body image existing in the human body detection result into a plurality of human body region pictures related to each human body image and stores the pictures;
step S32, the safety helmet wearing detection system carries out safety helmet detection on each human body area picture based on the safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture;
step S33, the safety helmet wearing detection system further calculates the color information of the image of the suspected safety helmet in the human body area image to obtain a color information calculation result corresponding to the suspected safety helmet area;
and step S34, the safety helmet wearing detection system combines the suspected safety helmet area identification result obtained in step S32 and the color information calculation result of the corresponding suspected safety helmet area obtained in step S33 to obtain a final detection result of the safety helmet, and the final detection result is output as the safety helmet detection result.
Preferably, the helmet wearing detection system in step S21 performs human body target detection on the monitored image by performing coordinate position detection on coordinates of each of the human body images existing in the monitored image.
Preferably, the helmet wearing detection system in step S22 performs human skeleton key point detection on each human body image, to perform coordinate position detection on coordinates of each human skeleton key point on the corresponding human body image.
Preferably, the step S33 of the helmet wearing detection system performing the picture color information calculation on the suspected helmet area in the human body area picture includes calculating contrast information, and/or chrominance information, and/or luminance information of each pixel point in the suspected helmet area.
The beneficial effects are that:
according to the method, the detection of the safety helmet is carried out after the detection of the human body, when the safety helmet is small, the safety helmet area can be extracted together through the detection of the human body, and the detection omission is reduced; meanwhile, the human body target detection and the human body skeleton key point detection are combined, so that the missing detection of the human body target when the human body is shielded can be reduced, the false detection when multiple persons are overlapped can be reduced, and the accuracy rate of the human body detection is improved; and detecting the safety helmet by using the target detection network, and improving the accuracy of safety helmet detection by combining the color information of the safety helmet area and the shape information of the safety helmet.
Drawings
FIG. 1 is a block diagram of a system for detecting wearing of safety helmets based on deep learning according to the present invention;
FIG. 2 is a block diagram of a safety helmet detection module provided in the present invention;
FIG. 3 is a step diagram of a method for detecting wearing of a safety helmet based on deep learning according to the present invention;
FIG. 4 is a flowchart illustrating an exemplary implementation of step S2 of FIG. 3;
FIG. 5 is a flowchart illustrating an exemplary implementation of step S3 of FIG. 3;
FIG. 6 is a block diagram illustrating the overall process of detecting a crash helmet by the crash helmet wear detection system of the present invention;
FIG. 7 is a block diagram of a process for detecting a human target by the system for detecting wearing of a safety helmet provided by the present invention;
FIG. 8 is a block diagram of a safety helmet detection system according to the present invention, after detecting a human target;
in the figure:
1. the safety helmet detection system comprises a data acquisition module, a human body detection result analysis module, a 4 image segmentation module, a 5 safety helmet detection module, a 6 early warning module, a 7 model training module, a 8 first safety helmet identification unit, a 9 second safety helmet identification unit, and a 10 safety helmet detection result forming unit.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.
Referring to fig. 1, the present invention provides a structure diagram of a helmet wearing detection system based on deep learning; the method comprises the following steps:
the data acquisition module 1 is used for acquiring a monitoring image at monitoring equipment;
a human body detection module 2 connected with the data acquisition module 1, wherein a human body target detection model and a human body skeleton key point detection model are preset in the human body detection module 2, and the human body detection module 2 is used for detecting a human body target of the monitored image based on the human body target detection model and obtaining a human body target detection result;
the human body detection module 2 is also used for detecting human body bone key points of a human body image existing in a human body target detection result based on the human body bone key point detection model and obtaining a human body bone key point detection result;
the human body detection result analysis module 3 is connected with the human body detection module 2 and is used for combining the human body target detection result and the human body bone key point detection result and obtaining a human body detection result through comprehensive analysis;
the image segmentation module 4 is connected with the human body detection result analysis module 3 and is used for segmenting each human body picture in the human body detection result into a plurality of human body region pictures related to each human body picture and storing the human body region pictures;
the safety helmet detection module 5 is connected with the image segmentation module 4 and used for identifying and detecting whether a safety helmet exists in each human body region picture according to a preset safety helmet detection model and obtaining a safety helmet detection result;
and the early warning module 6 is connected with the safety helmet detection module 5 and used for carrying out corresponding prompt warning according to the safety helmet detection result.
In the above technical solution, it should be noted that the human target detection model and the human skeleton key point detection model may be pre-trained models, and the human target detection model is obtained by training a deep convolutional neural network, where a network structure adopted by the deep convolutional neural network may be an existing YOLO neural network architecture, or may be other existing SSD or fast-RCNN neural network architectures. The human skeleton key point detection model is obtained through training of a deep convolution neural network, and the network structure adopted by the deep convolution neural network can be the existing Mask RCNN or FCN and other neural network architectures. The training processes of the human target detection model and the human skeleton key point detection model are the conventional convolutional neural network structure, and the training processes of the two models are not within the scope of the claims of the invention, so the training processes of the two models are not described in detail herein.
In addition, the safety helmet detection model is obtained by training a deep convolutional neural network, and the network structure of the deep convolutional neural network can adopt the existing lightweight network structure such as false-RCNN + MobileNet or tiny-yolo.
Preferably, the detection system for wearing safety helmet provided by the invention further comprises:
and the model training module 7 is respectively connected with the human body detection module 2 and the safety helmet detection module 5 and is used for training to form a human body target detection model, a human body skeleton key point detection model and a safety helmet detection model, updating the target detection model and the human body skeleton key point detection model in the human body detection module 2 and updating the safety helmet detection model in the safety helmet detection module 5.
It should be emphasized here that, in addition to training the human target detection model, the human bone key point detection model and the safety helmet detection model, the model training module 7 has another main function of automatically updating the newly trained model into the corresponding human detection module 2 or safety helmet detection module 5. For example, the images of the human body regions storing the image information of the safety helmet in the safety helmet detection result are input into the model training module 7, the model training module 7 finally trains to form a new safety helmet detection model according to a preset training method, such as applying a tiny-yolo deep convolution neural network, and the model training module 7 automatically updates the newly trained model into the safety helmet detection module 5 after the new safety helmet detection model is trained. That is to say, safety helmet detection model, human target detection model and human skeleton key point detection model are not fixed unchangeable, and the system will train the renewal to each model again automatically to improve this system's safety helmet and wear the rate of accuracy that detects.
Fig. 2 is a structural diagram of a safety helmet detection module provided in the present invention, and referring to fig. 2, a safety helmet detection module 5 specifically includes:
the first safety helmet identification unit 8 is used for carrying out safety helmet detection on each human body area picture based on a safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture;
the second safety helmet identification unit 9 is connected with the first safety helmet identification unit 8 and is used for further calculating the color information of the image of the suspected safety helmet in the human body area image to obtain a color information calculation result corresponding to the image of the suspected safety helmet area;
and the safety helmet detection result forming unit 10 is connected with the first safety helmet identification unit 8 and the second safety helmet identification unit 9, and is used for integrating the suspected safety helmet area identification result obtained by the identification of the first safety helmet identification unit 8 and the color information calculation result of the corresponding suspected safety helmet area obtained by the identification of the second safety helmet identification unit 9 to obtain a final detection result of the safety helmet, and the final detection result is used as a safety helmet detection result and is output.
It should be noted here that the picture color information calculated by the second helmet identification unit 9 includes contrast information, and/or chrominance information, and/or luminance information calculation associated with the corresponding human body region picture. The invention can easily judge whether the safety helmet exists in the human body area picture by calculating the picture color information corresponding to each human body area picture and comparing the picture color information with the preset standard picture color information corresponding to the safety helmet through picture information data, and can preliminarily confirm whether the safety helmet exists in the human body area picture.
However, sometimes, if the color of the clothes worn by the worker is completely the same as the color of the safety helmet, under such a situation, it is difficult for the system to judge whether the safety helmet exists in the human body region picture by judging the picture color information of the human body region picture. In order to solve the problem, the invention further provides a safety helmet detection method, namely, suspected safety helmet shape detection is carried out on each human body area picture, and image data comparison is carried out on the detected suspected safety helmet shape and a preset safety helmet standard shape image, so that whether a safety helmet exists in the human body area picture is finally confirmed.
The system finally forms the safety helmet detection result of the human body area picture according to the picture color information detection result of the human body area picture, namely the first safety helmet detection result, and according to the safety helmet shape detection result of the human body area picture, namely the second safety helmet detection result.
It should be noted here that the process of forming the final detection result of the safety helmet may include the following ways:
the system confirms that a safety helmet exists in a corresponding human body region picture according to a picture color information detection result, and confirms that a safety helmet exists in the corresponding human body region picture according to a safety helmet shape detection result, the system judges that the safety helmet exists in the human body region picture, and otherwise, the system judges that the safety helmet does not exist in the human body region picture.
And secondly, if the system confirms that the corresponding human body region picture has the safety helmet according to the picture color information detection result, but does not find that the safety helmet exists in the human body region picture according to the safety helmet shape detection result, the system judges that the safety helmet does not exist in the human body region picture.
And thirdly, if the system confirms that no safety helmet exists in the corresponding human body region picture according to the picture color information detection result, but confirms that a safety helmet exists in the human body region picture according to the safety helmet shape detection result, the system also judges that no safety helmet exists in the human body region picture.
The safety helmet detection method of the system can ensure the accuracy of safety helmet detection and avoid the occurrence of false detection as far as possible.
In addition, the above-mentioned human body target detection is to perform coordinate position detection on specific coordinates of the human body image existing in the monitored image. The method for detecting the coordinate position of the human body image is a coordinate positioning method commonly used in the prior art, and the coordinate positioning method is not the scope of the claimed invention, so the system does not describe the method for positioning the coordinate position of the human body image.
Since the structures and shapes of human body parts such as the brain, hands and feet and the like are different, the system detects key points of each human skeleton on the detected human body image so as to finally detect whether the safety helmet exists on the human body image. Specifically, the system detects the human skeleton key points by detecting the coordinate positions of the coordinates of the human skeleton key points on the corresponding human body image. It should be noted that the method adopted by the system for locating the coordinates of the key points of the human skeleton is a coordinate locating method existing in the prior art, and the coordinate locating method is not the scope of the claimed invention, so the method for locating the coordinates of the key points of the human skeleton by the system is not described herein.
The invention also provides a safety helmet wearing detection method based on deep learning, which is realized by the safety helmet wearing detection system, please refer to fig. 3, and specifically comprises the following steps:
step S1, the safety helmet wearing detection system acquires a monitoring image at the monitoring equipment;
step S2, the safety helmet wearing detection system carries out human body target detection on the monitored image and obtains a human body detection result;
step S3, the safety helmet wearing detection system performs safety helmet identification detection on each human body image existing in the human body target detection result to determine whether a safety helmet exists in each human body image, and obtains a safety helmet detection result.
Referring to the flowchart of fig. 6, firstly, a monitoring device collects a required human body image, then detects a human body and bone key points of the human body image by using a target detection network and a bone key point detection network through human body detection, and corrects a detection result by integrating the detection result to obtain an accurate human body detection result; and then inputting the segmented human body image into a safety helmet detection flow, performing safety detection by using a target detection network, and calculating color information of a safety helmet area to obtain a final safety helmet wearing detection result, so as to obtain a safety helmet wearing detection result of a monitoring area.
Referring to fig. 4, step S2 specifically includes the following steps:
step S21, the safety helmet wearing detection system carries out human body target detection on the monitored image based on a preset human body target detection model, and obtains a human body target detection result;
step S22, the safety helmet wearing detection system detects human skeleton key points of each human image in the human target detection result based on a preset human skeleton key point detection model, and obtains a human skeleton key point detection result;
and step S23, the safety helmet wearing detection system obtains a human body detection result through comprehensive analysis according to the human body target detection result and the human body bone key point detection result. ,
in the above technical solution, it should be noted that, the method for comprehensively obtaining the human body detection result based on the human body target detection result and the human body bone key point detection result by the safety helmet wearing detection system is as follows:
referring to fig. 7, when the system detects that there is a human body image, i.e. there is a human body target in the monitored image, the system further performs human body bone key point detection on the human body image. Because the shapes and appearances of main parts of a human body, such as the head and the hands and feet, are different, the system can detect whether corresponding human bones exist in the human body image according to the shape and the size of a preset human body standard head and the shape and the size corresponding to key points of each human skeleton, such as the hands and the feet. Meanwhile, because the positions of key parts of the human body such as the head, the hands and the feet on the human body are different, the system can perform coordinate positioning on the detected key points of the skeleton of the human body according to the existing coordinate positioning method, and further judge whether the human body image detected by the system is a real human body image or not, so that the accuracy of human body image detection is further improved.
Referring to fig. 5, step S3 further includes the following steps:
step S31, the safety helmet wearing detection system divides each human body image in the human body detection result into a plurality of human body area pictures related to each human body image and stores the pictures;
step S32, the safety helmet wearing detection system carries out safety helmet detection on each human body area picture based on the safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture;
step S33, the safety helmet wearing detection system further calculates the picture color information of the area where the suspected safety helmet is located in the human area picture to obtain a color information calculation result corresponding to the suspected safety helmet area;
in step S34, the safety helmet wearing detection system combines the suspected safety helmet area identification result obtained in step S32 and the color information calculation result of the corresponding suspected safety helmet area obtained in step S33 to obtain a final detection result for the safety helmet, and outputs the final detection result as a safety helmet detection result.
As described above, the detection of the human body target in the monitoring image by the headgear wearing detection system in step S21 is preferably performed by detecting the coordinate position of the coordinates of each human body image existing in the monitoring image. The coordinate position detection method is a coordinate positioning method existing in the prior art, and a specific positioning process of the coordinate position detection method is not described herein.
Further, the safety helmet wearing detection system in step S22 performs human skeleton key point detection on each human body image, that is, performs coordinate position detection on the coordinates of each human skeleton key point on the corresponding human body image. The detection method adopted by the coordinate position detection is also a coordinate positioning method existing in the prior art, and the specific positioning process is not described herein.
Referring to fig. 8, after the system performs suspected helmet area detection on a single human body area image, a suspected helmet area detection result and a detection accuracy score are obtained, and then according to the detection score, the color information of the image of the suspected helmet area is further detected, so as to finally obtain a helmet detection result.
It should be noted that, the detection accuracy score obtained by the system after detecting the suspected helmet area on each single human body area image is obtained by the following method:
the system scores the detection result of the suspected safety helmet area according to a preset evaluation method, for example, when the system detects that the detection probability of the suspected safety helmet area in the human body area image is greater than 0.3, the suspected safety helmet exists in the human body area image, otherwise, the suspected safety helmet does not exist in the human body area image; and finally, comprehensively judging whether the person in the image of the human body area correctly wears the safety helmet or not by combining the detection probability and the color information of the suspected safety helmet area.
Further, the step S33 of the helmet wearing detection system performing picture color information calculation on the suspected helmet area in the picture of the human area includes contrast information calculation, and/or chrominance information calculation, and/or luminance information calculation on each pixel point in the suspected helmet area.
In conclusion, the safety helmet detection is carried out after the human body target is detected, so that when the safety helmet is small, the safety helmet areas can be extracted together through the human body target detection, and the missing detection is reduced; meanwhile, the human body target detection and the bone key point detection are combined, so that missing detection when a human body is shielded can be reduced, false detection when multiple people are overlapped can be reduced, and the accuracy rate of human body target detection is improved; and simultaneously, the safety helmet is detected by using the target detection network, and the image color information and the safety helmet shape information of the safety helmet area are combined to detect the safety helmet, so that the accuracy of safety helmet detection is improved.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.
Claims (10)
1. A helmet wearing detection system based on deep learning is characterized by comprising:
the data acquisition module is used for acquiring a monitoring image at the monitoring equipment;
a human body detection module connected with the data acquisition module, wherein a human body target detection model and a human body skeleton key point detection model are preset in the human body detection module, and the human body detection module is used for detecting a human body target of the monitored image based on the human body target detection model and obtaining a human body target detection result;
the human body detection module is also used for detecting human body bone key points of a human body image existing in the human body target detection result based on the human body bone key point detection model and obtaining a human body bone key point detection result;
the human body detection result analysis module is connected with the human body detection module and is used for combining the human body target detection result and the human body bone key point detection result and obtaining a human body detection result through comprehensive analysis;
the image segmentation module is connected with the human body detection result analysis module and is used for segmenting each human body image in the human body detection result into a plurality of human body region pictures related to each human body image and storing the human body region pictures;
the safety helmet detection module is connected with the image segmentation module and used for identifying and detecting whether a safety helmet exists in each human body region picture according to a preset safety helmet detection model and obtaining a safety helmet detection result;
and the early warning module is connected with the safety helmet detection module and used for carrying out corresponding prompt warning according to the safety helmet detection result.
2. The system of claim 1, further comprising a model training module, respectively connected to the human body detection module and the safety helmet detection module, for training to form the human body target detection model, the human body bone key point detection model and the safety helmet detection model, and updating the target detection model and the human body bone key point detection model in the human body detection module and the safety helmet detection model in the safety helmet detection module.
3. The system for detecting the wearing of a safety helmet according to claim 1, wherein the safety helmet detection module specifically comprises:
the first safety helmet identification unit is used for carrying out safety helmet detection on each human body area picture based on the safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture or not;
the second safety helmet identification unit is connected with the first safety helmet identification unit and used for further calculating the color information of the image of the area where the suspected safety helmet is located in the human body area image to obtain a color information calculation result corresponding to the image of the area where the suspected safety helmet is located;
and the safety helmet detection result forming unit is connected with the first safety helmet identification unit and the second safety helmet identification unit and used for integrating the suspected safety helmet area identification result obtained by the first safety helmet identification unit and the color information calculation result corresponding to the suspected safety helmet area obtained by the second safety helmet identification unit to obtain a final detection result of the safety helmet as the safety helmet detection result and outputting the final detection result.
4. The headgear wear detection system of claim 1, wherein the human target detection is coordinate position detection of coordinates of each of the human images in the monitor image, and the human bone key detection is coordinate position detection of coordinates of each of the human bone key points on each of the human images on the corresponding human image.
5. A safety helmet wearing detection method based on deep learning is realized by applying the safety helmet wearing detection system as claimed in any one of claims 1 to 4, and is characterized by comprising the following steps:
step S1, the safety helmet wearing detection system obtains the monitoring image at the monitoring device;
step S2, the safety helmet wearing detection system carries out human body target detection on the monitoring image and obtains the human body target detection result;
step S3, the safety helmet wearing detection system performs safety helmet identification detection on each human body image in the human body target detection result to determine whether a safety helmet exists in each human body image, and obtains the safety helmet detection result.
6. The method for detecting wearing of a safety helmet according to claim 5, wherein the step S2 specifically includes the steps of:
step S21, the safety helmet wearing detection system carries out human body target detection on the monitoring image based on a preset human body target detection model, and obtains the human body target detection result;
step S22, the safety helmet wearing detection system detects human skeleton key points of each human body image in the human body target detection result based on a preset human skeleton key point detection model, and obtains the human skeleton key point detection result;
and step S23, the safety helmet wearing detection system obtains the human body detection result through comprehensive analysis according to the human body target detection result and the human body bone key point detection result.
7. The method for detecting wearing of a safety helmet according to claim 5, wherein the step S3 specifically includes the steps of:
step S31, the headgear wearing detection system divides each human body image existing in the human body detection result into a plurality of human body region pictures related to each human body image and stores the pictures;
step S32, the safety helmet wearing detection system carries out safety helmet detection on each human body area picture based on the safety helmet detection model so as to identify whether a suspected safety helmet exists in each human body area picture;
step S33, the safety helmet wearing detection system further calculates the color information of the image of the suspected safety helmet in the human body area image to obtain a color information calculation result corresponding to the suspected safety helmet area;
step S34, the safety helmet wearing detection system combines the suspected safety helmet area identification result obtained in step S32 and the color information calculation result corresponding to the suspected safety helmet area obtained in step S33 to obtain a final detection result for a safety helmet, and outputs the final detection result as the safety helmet detection result.
8. The headgear wearing detection method according to claim 6, wherein the detecting of the human object in the monitored image by the headgear wearing detection system in step S21 is detecting a coordinate position of coordinates in the monitored image of each of the human images existing in the monitored image.
9. The method for detecting wearing of a helmet according to claim 6, wherein the detecting system for wearing a helmet in step S22 performs human skeleton key point detection on each human body image to perform coordinate position detection on coordinates of each human skeleton key point on the corresponding human body image.
10. The method for detecting wearing of safety helmet according to claim 7, wherein the step S33 of the system for detecting wearing of safety helmet in the picture of human body area performing the color information calculation on the suspected safety helmet area includes calculating contrast information, and/or chrominance information, and/or luminance information of each pixel point in the suspected safety helmet area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911190920.3A CN110889376A (en) | 2019-11-28 | 2019-11-28 | Safety helmet wearing detection system and method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911190920.3A CN110889376A (en) | 2019-11-28 | 2019-11-28 | Safety helmet wearing detection system and method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110889376A true CN110889376A (en) | 2020-03-17 |
Family
ID=69749206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911190920.3A Pending CN110889376A (en) | 2019-11-28 | 2019-11-28 | Safety helmet wearing detection system and method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110889376A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111428670A (en) * | 2020-03-31 | 2020-07-17 | 南京甄视智能科技有限公司 | Face detection method, face detection device, storage medium and equipment |
| CN111476117A (en) * | 2020-03-25 | 2020-07-31 | 中建科技有限公司深圳分公司 | Safety helmet wearing detection method and device and terminal |
| CN112257580A (en) * | 2020-10-21 | 2021-01-22 | 中国石油大学(华东) | Human body key point positioning detection method based on deep learning |
| CN112364778A (en) * | 2020-11-12 | 2021-02-12 | 上海明华电力科技有限公司 | Power plant safety behavior information automatic detection method based on deep learning |
| CN112581525A (en) * | 2020-12-03 | 2021-03-30 | 郑州捷安高科股份有限公司 | Method, device and equipment for detecting state of human body wearing article and storage medium |
| CN112990149A (en) * | 2021-05-08 | 2021-06-18 | 创新奇智(北京)科技有限公司 | Multi-mode-based high-altitude safety belt detection method, device, equipment and storage medium |
| CN113139426A (en) * | 2021-03-12 | 2021-07-20 | 浙江智慧视频安防创新中心有限公司 | Detection method and device for wearing safety helmet, storage medium and terminal |
| CN113536842A (en) * | 2020-04-15 | 2021-10-22 | 普天信息技术有限公司 | Electric power operator safety dressing identification method and device |
| WO2022213552A1 (en) * | 2021-04-07 | 2022-10-13 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for helmet detection |
| CN116311082A (en) * | 2023-05-15 | 2023-06-23 | 广东电网有限责任公司湛江供电局 | Wearing detection method and system based on matching of key parts and images |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130286197A1 (en) * | 2012-04-27 | 2013-10-31 | Hon Hai Precision Industry Co., Ltd. | Safety guard device and method for detecting falling object |
| CN108038469A (en) * | 2017-12-27 | 2018-05-15 | 百度在线网络技术（北京）有限公司 | Method and apparatus for detecting human body |
| CN108288033A (en) * | 2018-01-05 | 2018-07-17 | 电子科技大学 | A kind of safety cap detection method merging multiple features based on random fern |
| CN108319934A (en) * | 2018-03-20 | 2018-07-24 | 武汉倍特威视系统有限公司 | Safety cap wear condition detection method based on video stream data |
| CN108647619A (en) * | 2018-05-02 | 2018-10-12 | 安徽大学 | The detection method and device that safety cap is worn in a kind of video based on deep learning |
| CN109272347A (en) * | 2018-08-16 | 2019-01-25 | 苏宁易购集团股份有限公司 | A kind of statistical analysis technique and system of shops's volume of the flow of passengers |
| CN110046600A (en) * | 2019-04-24 | 2019-07-23 | 北京京东尚科信息技术有限公司 | Method and apparatus for human testing |
-
2019
- 2019-11-28 CN CN201911190920.3A patent/CN110889376A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130286197A1 (en) * | 2012-04-27 | 2013-10-31 | Hon Hai Precision Industry Co., Ltd. | Safety guard device and method for detecting falling object |
| CN108038469A (en) * | 2017-12-27 | 2018-05-15 | 百度在线网络技术（北京）有限公司 | Method and apparatus for detecting human body |
| CN108288033A (en) * | 2018-01-05 | 2018-07-17 | 电子科技大学 | A kind of safety cap detection method merging multiple features based on random fern |
| CN108319934A (en) * | 2018-03-20 | 2018-07-24 | 武汉倍特威视系统有限公司 | Safety cap wear condition detection method based on video stream data |
| CN108647619A (en) * | 2018-05-02 | 2018-10-12 | 安徽大学 | The detection method and device that safety cap is worn in a kind of video based on deep learning |
| CN109272347A (en) * | 2018-08-16 | 2019-01-25 | 苏宁易购集团股份有限公司 | A kind of statistical analysis technique and system of shops's volume of the flow of passengers |
| CN110046600A (en) * | 2019-04-24 | 2019-07-23 | 北京京东尚科信息技术有限公司 | Method and apparatus for human testing |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111476117A (en) * | 2020-03-25 | 2020-07-31 | 中建科技有限公司深圳分公司 | Safety helmet wearing detection method and device and terminal |
| CN111428670A (en) * | 2020-03-31 | 2020-07-17 | 南京甄视智能科技有限公司 | Face detection method, face detection device, storage medium and equipment |
| CN111428670B (en) * | 2020-03-31 | 2022-08-09 | 南京甄视智能科技有限公司 | Face detection method, face detection device, storage medium and equipment |
| CN113536842A (en) * | 2020-04-15 | 2021-10-22 | 普天信息技术有限公司 | Electric power operator safety dressing identification method and device |
| CN112257580A (en) * | 2020-10-21 | 2021-01-22 | 中国石油大学(华东) | Human body key point positioning detection method based on deep learning |
| CN112364778A (en) * | 2020-11-12 | 2021-02-12 | 上海明华电力科技有限公司 | Power plant safety behavior information automatic detection method based on deep learning |
| CN112581525A (en) * | 2020-12-03 | 2021-03-30 | 郑州捷安高科股份有限公司 | Method, device and equipment for detecting state of human body wearing article and storage medium |
| CN112581525B (en) * | 2020-12-03 | 2023-02-28 | 郑州捷安高科股份有限公司 | Method, device and equipment for detecting state of human body wearing article and storage medium |
| CN113139426A (en) * | 2021-03-12 | 2021-07-20 | 浙江智慧视频安防创新中心有限公司 | Detection method and device for wearing safety helmet, storage medium and terminal |
| WO2022213552A1 (en) * | 2021-04-07 | 2022-10-13 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for helmet detection |
| CN112990149A (en) * | 2021-05-08 | 2021-06-18 | 创新奇智(北京)科技有限公司 | Multi-mode-based high-altitude safety belt detection method, device, equipment and storage medium |
| CN116311082A (en) * | 2023-05-15 | 2023-06-23 | 广东电网有限责任公司湛江供电局 | Wearing detection method and system based on matching of key parts and images |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110889376A (en) | Safety helmet wearing detection system and method based on deep learning | |
| CN111445524B (en) | Scene understanding-based construction site worker unsafe behavior identification method | |
| US20200237266A1 (en) | Action recognition and pose estimation method and apparatus | |
| CN109034215A (en) | A kind of safety cap wearing detection method based on depth convolutional neural networks | |
| CN115294533B (en) | Building construction state monitoring method based on data processing | |
| CN109389040B (en) | Inspection method and device for safety dressing of personnel in operation field | |
| CN112270807A (en) | Old man early warning system that tumbles | |
| CN112800901A (en) | Mine personnel safety detection method based on visual perception | |
| CN111325133A (en) | Image processing system based on artificial intelligence recognition | |
| CN114155492A (en) | High-altitude operation safety belt hanging rope high-hanging low-hanging use identification method and device and electronic equipment | |
| CN115797856A (en) | Intelligent construction scene safety monitoring method based on machine vision | |
| CN112597903B (en) | Electric power personnel safety state intelligent identification method and medium based on stride measurement | |
| CN114155557A (en) | Positioning method, positioning device, robot and computer-readable storage medium | |
| CN117853979A (en) | Method and device for detecting wearing of safety helmet, electronic equipment and storage medium | |
| CN113221815A (en) | Gait identification method based on automatic detection technology of skeletal key points | |
| CN116965781B (en) | Method and system for monitoring vital signs and driving behaviors of driver | |
| CN111291728A (en) | Detection system, detection equipment and detection method for illegal crossing of transmission belt behavior | |
| CN116884088A (en) | Monitoring scene sleep sentry recognition method | |
| CN111274888B (en) | Helmet and work clothes intelligent identification method based on wearable mobile glasses | |
| CN116229502A (en) | Image-based tumbling behavior identification method and equipment | |
| KR20230099225A (en) | An apparatus for monitoring safety status of construction workers in an image and operating method thereof | |
| CN114998778A (en) | Wearing compliance detection method, detection device and computer readable storage medium | |
| CN116645782B (en) | Safety helmet belt detection method based on image recognition | |
| CN113033239B (en) | Behavior detection method and device | |
| CN112800900A (en) | Mine personnel land falling detection method based on visual perception |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200317 |
|
| WD01 | Invention patent application deemed withdrawn after publication |