Image-based wearable device detection system and method thereof

Abstract

The invention relates to the technical field of computer vision application, and provides an image-based wearable device recognition system, which comprises an acquisition unit used for acquiring a detection picture; the first judgment unit is used for judging whether a human body exists in the detection picture; the face recognition unit is used for performing face recognition on the human body to obtain ID information corresponding to the human body; the first determining unit is used for determining wearable equipment needing to be worn by the human body according to the human body ID information; the second judgment unit is used for judging whether the wearable device needing to be worn by the human body exists in the detection picture or not; the second determining unit is used for determining a correct wearing area according to the posture of the human body if the detected picture has the wearable equipment needing to be worn; the third judgment unit is used for judging whether the wearable equipment is in a correct wearing area or not; and the alarm unit is used for judging that the human body correctly wears the wearable equipment if the wearable equipment needing to be worn is in the correct wearing area, and executing alarm operation if the wearable equipment is not in the correct wearing area.

Description

Technical Field

[0001] The present invention relates to the field of computer vision application technology, and in particular to an image-based wearable device detection system and method thereof. Background Art

[0002] Most safety accidents at production construction sites are related to workers violating regulations. Falling, hitting objects, collapse and hoisting damage are the four major types of fatal accidents. Moreover, falling and being hit by objects are the most common types of fatal accidents. However, due to the wide geographical distribution of construction sites, numerous equipment, complex environment, large amount of operations, and limited safety supervision, it is difficult to achieve real-time and full-process safety management of production construction sites.

[0003] In the field of safety wearable device detection, Chinese patent: CN109670441A discloses a method for realizing helmet wearing recognition, comprising the following steps: obtaining a detection image; judging whether there is a human body and a safety helmet in the detection image; if there is a human body and a safety helmet in the detection image, obtaining the posture of the human body, and determining the correct wearing area according to the posture of the human body; judging whether the safety helmet is in the correct wearing area; if so, determining that the human body is wearing the safety helmet correctly, if not, performing an alarm operation. However, in the above-mentioned safety wearable device detection method, there are the following shortcomings: First, this method can only monitor the safety helmet, but in addition to the safety helmet, which is a necessary safety protection equipment on the construction site, safety ropes and gloves are also indispensable protection equipment. Second, the image acquisition device of this method is limited to one location (entrance and exit), and it is impossible to shoot the entire construction site, and it is impossible to perform mobility detection on the staff used in the construction site. When the staff accidentally removes the equipment during work and forgets to put it on again, it is impossible to notify the staff to wear safety equipment in time. 3. Given that there are many types of work on construction sites, different types of work require different installation wearable devices. This method cannot obtain the identity information of the staff through face recognition, and judge whether the staff is wearing the corresponding wearable device correctly based on the identity information of the staff, thereby issuing different alarm messages according to different wearing conditions. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention proposes the following technical solutions:

[0005] In the first aspect, the present invention provides an image-based wearable device recognition system, comprising: an acquisition unit for acquiring a detection image; a first judgment unit for judging whether there is a human body in the detection image; a face recognition unit for performing face recognition on the human body to obtain ID information corresponding to the human body; a first determination unit for determining a wearable device that the human body needs to wear according to the human body ID information; a second judgment unit for judging whether there is a wearable device that the human body needs to wear in the detection image; a second determination unit for acquiring the posture of the human body if the wearable device that the human body needs to wear exists in the detection image, and determining a correct wearing area according to the posture of the human body; a third judgment unit for judging whether the wearable device that the human body needs to wear is in the correct wearing area; an alarm unit for judging whether the wearable device that the human body needs to wear is correctly worn if the wearable device that the human body needs to wear is in the correct wearing area, and performing an alarm operation if the wearable device that the human body needs to wear is not in the correct wearing area.

[0006] A second aspect: The present invention provides an image-based wearable device recognition method, comprising the following steps:

[0007] S1, obtain the detection picture;

[0008] S2, determining whether there is a human body in the detection image;

[0009] S3, if there is a human body in the detection image, perform face recognition on the human body to obtain ID information corresponding to the human body, and obtain a wearable device that the human body needs to wear according to the ID information;

[0010] S4, determining whether the wearable device to be worn exists in the detection image;

[0011] S5, if the detection image contains the wearable device to be worn, obtaining the posture of the human body, and determining the correct wearing area according to the posture of the human body;

[0012] S6, determining whether the wearable device to be worn is in the correct wearing area;

[0013] S7: If yes, determine whether the wearable device is worn correctly by the human body; if no, perform an alarm operation.

[0014] Preferably, step S1 includes: acquiring a real-time video stream through a camera, parsing the real-time video stream to obtain a picture to be analyzed, wherein the picture to be analyzed includes an intra-coded frame and a predictive coded frame, and obtaining an optical flow component value of the picture to be analyzed; if the optical flow component value of the picture to be analyzed is greater than a first preset threshold, selecting the picture to be analyzed as a detection picture.

[0015] Preferably, step S2 includes: collecting human image samples and non-human image samples, classifying, labeling and training them, and obtaining a human feature model; using k different rectangular frames to nominate on the convolutional feature layer of the detection image, and screening out candidate areas where human bodies may exist, where k is a positive integer; extracting human features from the candidate areas respectively; using the human feature model to classify and regress the features, and identify whether there is a human body in the detection image; if it is identified that there is a human body in the detection image, performing border position regression processing on the human body to obtain a human body frame.

[0016] Preferably, the step S3 includes: storing the correspondence between the ID information of the human body and the wearable device to be worn, and when the ID information of the human body is identified, directly acquiring one or more wearable devices to be worn by the human body according to the correspondence.

[0017] Preferably, step S4 includes: collecting a variety of wearable device image samples and non-wearable device image samples, classifying, labeling and training, and obtaining a variety of wearable device models; using k different rectangular boxes on the convolutional feature layer of the detection image to nominate and screen out candidate areas where wearable devices may exist, where k is a positive integer; extracting features of wearable devices from the candidate areas respectively; using the wearable device model to classify and regress the features to identify whether the detection image has the wearable device to be worn; if it is identified that the detection image has the wearable device to be worn, performing border position regression processing on the wearable device to be worn to obtain a wearable device frame.

[0018] Preferably, obtaining the posture of the human body includes: obtaining the aspect ratio α of the human body frame; if the value of α is greater than a second preset threshold, determining that the posture of the human body is standing; if the value of α is less than the second preset threshold, determining that the posture of the human body is squatting.

[0019] Preferably, the wearable device includes a safety helmet, a safety rope and gloves, and the determining of the correct wearing area according to the posture of the human body includes: if the posture of the human body is standing, determining the correct wearing area as a first safety helmet wearing area and / or a first safety rope wearing area and / or a first glove wearing area according to the ID information, the first safety helmet wearing area being a square area greater than a first preset height at the upper part of the human body frame; the first safety rope wearing area being a square area from a second preset height to a first preset height in the middle part of the human body frame, and the first glove wearing area being a square area from a third preset height to a second preset height in the lower part of the human body frame; if the posture of the human body is squatting, determining the correct wearing area as a second safety helmet wearing area and / or a second safety rope wearing area and / or a second glove wearing area according to the ID information, the second safety helmet wearing area being a square area greater than a fourth preset height at the upper part of the human body frame; the second safety rope wearing area being a square area from a fifth preset height to a fourth preset height in the middle part of the human body frame, and the second glove wearing area being a square area below the fifth preset height in the lower part of the human body frame.

[0020] Preferably, the S6 step includes: obtaining a wearable device that has an overlapping area with the human body frame and the overlapping area is within the head, torso, and hand of the human body as a target wearable device; if the posture of the human body is standing, judging whether the wearable device frame of the target wearable device is respectively located in the first safety helmet wearing area, the first safety rope wearing area, and the first glove wearing area; if so, judging that the target wearable device is within the correct wearing area; if the posture of the human body is squatting, judging whether the wearable device frame of the target wearable device is respectively located in the second safety helmet wearing area, the second safety rope wearing area, and the second glove wearing area; if so, judging that the target wearable device is within the correct wearing area.

[0021] Preferably, in the step S7: when it is detected that a human body is not wearing the wearable device correctly, the ID information of the human body and the wearable device information corresponding to the ID information are broadcasted.

[0022] The beneficial effects achieved by the present invention are:

[0023] (1) It can conduct all-round detection on wearable devices such as safety helmets, safety ropes, gloves, etc. Different types of work require different wearable devices. The present invention can identify the ID information of the worker, and determine the type of work of the worker based on the ID information, and detect whether the worker is wearing the wearable device correctly based on the type of work. If the worker does not wear the wearable device correctly, the identity information of the worker and the wearable device that the worker should wear are broadcasted.

[0024] (2) Multiple cameras are installed throughout the construction site to capture all-round footage of the entire construction site. When a worker accidentally removes his / her equipment during work and forgets to put it back on, the worker is promptly notified to wear safety equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate the same parts.

[0026] Figure 1 A flow chart of a wearable device identification method based on an image provided by an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of an image-based wearable device recognition system provided by another embodiment of the present invention;

[0028] Figure 3 A schematic diagram of the detection model training process provided by an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of a specific implementation flow of step S101 in an embodiment of the present invention;

[0030] Figure 5 This is a schematic diagram of a specific implementation process of steps S102 and S104 of an embodiment of the present invention;

[0031] Figure 6 It is a schematic diagram of a specific implementation flow of steps S105-S107 of the present invention. DETAILED DESCRIPTION

[0032] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0033] See also Figure 1 , which is a flow chart of a wearable device identification method based on an image provided by an embodiment of the present invention,

[0034] As can be seen from the figure, it includes the following steps:

[0035] S101, obtaining a detection image;

[0036] Multiple cameras are set up on the construction site to capture the situation inside the construction site in real time from multiple angles. When a worker accidentally takes off the wearable device during work and forgets to put it on again thereafter, one or more of the multiple cameras can obtain relevant videos and perform relevant identification and detection. When it is found that the worker is wearing the wearable device correctly, the worker is promptly notified to wear the wearable device.

[0037] See also Figure 4 ,The camera obtains real-time video stream, and the processor parses the real-time video stream to obtain the picture to be analyzed.

[0038] Generally, there are three types of coded frames in a video stream: intra-coded frames (I frames), predictive coded frames (P frames), and bidirectional coded frames (B frames). I frames use only the spatial correlation of the frame for video compression, P frames use forward reference frames for temporal domain predictive coding, and B frames use forward and backward bidirectional reference frames for temporal domain predictive coding. Usually, I frames have a low compression ratio, slightly better image quality, and are the basis of P frames and B frames, so I frames should be selected for analysis first, followed by P frames that contain more information. Since the time difference between two adjacent I frames is relatively long, several P frames should also be selected between two adjacent I frames as images to be analyzed to prevent key information from being missed.

[0039] Obtain the optical flow component value of the image to be analyzed.

[0040] This example uses the change of pixels in the image sequence in the time domain and the correlation between adjacent frames to find the corresponding relationship between the previous frame and the current frame, thereby calculating the motion information of the object between adjacent frames. The optical flow field is to find the movement speed and direction of each pixel in the image through a picture sequence. This embodiment obtains the optical flow component value of the picture to be analyzed by using the sparse optical flow calculation method:

[0041]

[0042] Wherein, is the optical flow component value, which indicates the motion offset of the corner point in the x direction. The larger the value, the larger the motion offset in the x direction, and the larger the difference between the two frames of images. If the optical flow component value of the image to be analyzed is greater than the first preset threshold, the image to be analyzed is selected as the detection image.

[0043] In the specific implementation, the first preset threshold is set to 1.2. If the optical flow component value of the image to be analyzed is greater than 1.2, it is determined that the difference between the image to be analyzed and the previous frame image is large. In order to avoid omissions in the detection results, the image to be analyzed should be selected as the detection image for the next step of analysis.

[0044] It should be noted that those skilled in the art can also determine the difference between the image to be analyzed and the previous frame image by obtaining the motion offset of the image in the y direction, and select the image to be analyzed by themselves, thereby improving both the accuracy and efficiency of detection.

[0045] S102, determining whether there is a human body in the detection image;

[0046] See also Figure 3 ,In the specific implementation, it is necessary to train the human ,feature model.

[0047] Collect human image samples and non-human image samples, classify, label and train them to obtain human feature models. Collect a large number of human image samples and non-human image samples, use special tools to label them, and divide them into training samples and test samples. The training samples are used to train the parameters of the model, and the test samples are used to test the effect of the model. This example uses 20,000 human image samples and non-human image samples, and uses the ImgLabel tool to label the areas and categories. Divide them into training set and test set in a ratio of 7:3.

[0048] See also Figure 5 , k different rectangular boxes are used to nominate on the convolutional feature layer of the detection image to screen out candidate areas where human bodies may exist, where k is a positive integer.

[0049] Extracting features of human body from the candidate regions respectively;

[0050] In the specific implementation, the extracted image features mainly include: image texture features, edge features and motion features. (1) Texture features mainly include the grayscale histogram, edge direction histogram, grayscale co-occurrence matrix, etc. (2) Edge features directly show the outline of the image, mainly including the perimeter, area, aspect ratio (principal axis ratio), dispersion, compactness, etc. (3) Motion features are related to motion behavior, generally including motion center of mass, speed, displacement, gradient, etc.

[0051] In the specific implementation, the entire image is input and the Darknet51 network is used to automatically obtain the features of the image.

[0052] The features are classified and regressed using a human feature model to identify whether the detection image contains a human body; if it is identified that the detection image contains a human body, a frame position regression process is performed on the human body to obtain a human body frame.

[0053] In the specific implementation, the Feature Map after feature extraction is classified into two categories to determine whether it contains a human body, and then k regression models (each corresponding to a different Anchor Box) are used to adjust the position and size of the candidate box, and finally the human body is classified.

[0054] In this example, the model further processes the predicted values ​​of the bounding boxes of each Feature in the Feature Map (the relative coordinates of the center point and the relative width and height of the box) to obtain the actual values. The predicted center coordinates are normalized using the sigmoid function, and the actual coordinates are obtained by adding the offset value of the current Feature relative to the upper left corner of the image. The predicted object width and height are normalized using the exp function, and then multiplied by the width and height of the prior box corresponding to the current Anchor Box to obtain the actual width and height. The width and height of the prior box corresponding to each Anchor Box can be obtained by clustering the data set.

[0055] This embodiment uses the detection and classification model and the GPU to determine whether each area contains a specific type of target.

[0056] The detection model is Darknet, and the classifier is the YOLOV3 model based on the focal loss function. The detected human body is then post-processed by border position regression to obtain the final human body frame.

[0057] S103: If a human body exists in the detection image, face recognition is performed on the human body to obtain ID information corresponding to the human body, and a wearable device that the human body needs to wear is obtained according to the ID information.

[0058] In a specific implementation, after a human body is detected, the ID information of the human body is further determined, and the wearable device that the human body needs to wear is determined based on the ID information of the human body.

[0059] When construction workers join the company, their ID information and the type of work they are responsible for are collected, and the wearable devices they need to wear are determined based on the type of work they are responsible for, and the ID information of the worker is associated with the wearable device they are wearing.

[0060] In a specific embodiment, by storing the correspondence between the ID information of the human body and the wearable device that the human body needs to wear, after identifying the ID information of the human body, the correspondence can be directly retrieved to obtain the wearable device that the human body needs to wear. The wearable devices include safety helmets, safety ropes, gloves, etc. In view of the fact that there are many types of work on the construction site, and different types of work have different requirements for wearable devices, it can be obtained through the corresponding relationship that the human body should wear one or more of the safety helmets, safety ropes, and gloves. The ID information of the human body is obtained by face recognition technology, wherein the face recognition technology adopts existing conventional technology, which will not be described in detail here.

[0061] S104, determining whether the wearable device to be worn exists in the detection image.

[0062] See also Figure 3In the specific implementation, it is necessary to train the wearable device feature model. Collect a variety of wearable device image samples and non-wearable device image samples, classify, label and train them, and obtain a variety of wearable device feature models.

[0063] Collect a large number of wearable device image samples and non-wearable device image samples, use special tools to annotate them, and divide them into training samples and test samples. The training samples are used to train the model parameters, and the test samples are used to test the model effect. This example uses 8,000 wearable device image samples and non-wearable device image samples, and uses the ImgLabel tool to annotate the area and category. Divide them into training set and test set in a ratio of 7:3.

[0064] See also Figure 5 , k different rectangular boxes are used to nominate on the convolutional feature layer of the detection image to screen out candidate areas where wearable devices may exist, where k is a positive integer.

[0065] Extracting features of wearable devices from the candidate areas respectively;

[0066] In the specific implementation, the extracted image features mainly include: image texture features, edge features and motion features. (1) Texture features mainly include the grayscale histogram, edge direction histogram, grayscale co-occurrence matrix, etc. (2) Edge features directly show the outline of the image, mainly including the perimeter, area, aspect ratio (principal axis ratio), dispersion, compactness, etc. (3) Motion features are related to motion behavior, generally including motion center of mass, speed, displacement, gradient, etc.

[0067] In the specific implementation, the entire image is input and the Darknet51 network is used to automatically obtain the features of the image.

[0068] The features are classified and regressed using a wearable device feature model to identify whether the detection image has a wearable device; if it is identified that the detection image has a wearable device to be worn, a frame position regression process is performed on the wearable device to be worn to obtain a wearable device frame.

[0069] In the specific implementation, the Feature Map after feature extraction is classified into two categories to determine whether it contains wearable devices, and then k regression models (each corresponding to a different Anchor Box) are used to adjust the position and size of the candidate box, and finally the wearable device classification is performed.

[0070] In this example, the model further processes the predicted values ​​of the bounding boxes of each Feature in the Feature Map (the relative coordinates of the center point and the relative width and height of the box) to obtain the actual values. The predicted center coordinates are normalized using the sigmoid function, and the actual coordinates are obtained by adding the offset value of the current Feature relative to the upper left corner of the image. The predicted object width and height are normalized using the exp function, and then multiplied by the width and height of the prior box corresponding to the current Anchor Box to obtain the actual width and height. The width and height of the prior box corresponding to each Anchor Box can be obtained by clustering the data set.

[0071] This embodiment uses the detection and classification model and GPU to determine whether each area contains a specific type of target. The detection model is Darknet, and the classifier is the YOLOV3 model based on the focal loss function. Then, the detected wearable device to be worn is processed by the frame position regression post-processing operation to obtain the final wearable device frame.

[0072] S105: If the detection image contains the wearable device to be worn, the posture of the human body is obtained, and a correct wearing area is determined according to the posture of the human body.

[0073] In a specific implementation, after detecting a human body and the wearable device that the human body needs to wear, it is necessary to further determine whether the human body is wearing the wearable device correctly. Due to the shooting angle, the size of the human body close to the lens and the human body far away from the lens are different, and the wearable device is also similar; therefore, the embodiment of the present invention simulates the data when the wearable device is correctly worn based on the acquired human body posture to determine whether the human body is wearing the wearable device correctly. The specific steps are as follows:

[0074] The height-to-width ratio α of the human body frame is obtained; if the value of α is greater than a second preset threshold, the posture of the human body is determined to be standing; if the value of α is less than the second preset threshold, the posture of the human body is determined to be squatting.

[0075] In a specific implementation, the posture of the human body is determined according to the aspect ratio of the human body frame. In this embodiment, the second preset threshold is set to 1.8. If the value of α is greater than 1.8, the posture of the human body is determined to be standing; if the value of α is less than 1.8, the posture of the human body is determined to be squatting; technical personnel in this field can also determine the value of the second preset threshold by themselves according to the shooting angle of the picture or other factors, and the present invention does not make specific limitations on this.

[0076] The wearable device includes a safety helmet, a safety rope and gloves. If the posture of the human body is standing, the correct wearing area is determined to be a first safety helmet wearing area and / or a first safety rope wearing area and / or a first glove wearing area, and the first safety helmet wearing area is a square area greater than a first preset height on the upper part of the human body frame; the first safety rope wearing area is a square area from a second preset height to a first preset height in the middle part of the human body frame, and the first glove wearing area is a square area from a third preset height to a second preset height in the lower part of the human body frame; if the posture of the human body is squatting, the correct wearing area is determined to be a second safety helmet wearing area and / or a second safety rope wearing area and / or a second glove wearing area, and the second safety helmet wearing area is a square area greater than a fourth preset height on the upper part of the human body frame; the second safety rope wearing area is a square area from a fifth preset height to a fourth preset height in the middle part of the human body frame, and the second glove wearing area is a square area below the fifth preset height in the lower part of the human body frame.

[0077] S106, determining whether the wearable device to be worn is within the correct wearing area.

[0078] Obtain a wearable device whose overlapping area with the human body frame and the overlapping area on the head, torso, and hand of the human body as the target wearable device; if the posture of the human body is standing, determine whether the wearable device frame of the target wearable device is respectively located in the first safety helmet wearing area, the first safety rope wearing area, and the first glove wearing area; if so, determine that the target wearable device is in the correct wearing area; if the posture of the human body is squatting, determine whether the wearable device frame of the target wearable device is respectively located in the second safety helmet wearing area, the second safety rope wearing area, and the second glove wearing area; if so, determine that the target wearable device is in the correct wearing area.

[0079] In a specific implementation, the target wearable device includes a target helmet, a target safety rope, and a target glove. After obtaining the target helmet, the target safety rope, and the target glove, based on the posture of the human body, if the posture of the human body is determined to be standing, it is determined whether the target helmet is located above the first preset height of the upper part of the human body frame, or whether the target safety rope is located between the second preset height and the first preset height in the middle part of the human body frame, or whether the target glove is located between the third preset height and the second preset height in the lower part of the human body frame. If so, it is determined that the target wearable device is in the correct wearing area.

[0080] If the human body posture is determined to be squatting, determine whether the target safety helmet is located above the fourth preset height of the upper part of the human body frame, or determine whether the target safety rope is located between the fifth preset height and the fourth preset height in the middle part of the human body frame, or determine whether the target gloves are located below the fifth preset height of the lower part of the human body frame. If so, determine that the target wearable device is in the correct wearing area.

[0081] S107: If yes, determine whether the wearable device is worn correctly by the human body; if no, perform an alarm operation.

[0082] In the specific implementation, if the wearable device is in the correct wearing area, it is determined that the person is wearing the wearable device correctly and there is no violation in the detection picture; if the wearable device is not in the correct wearing area, it is determined that the person is not wearing the wearable device correctly, there is a violation in the detection picture, and an alarm operation is performed. The alarm method is to broadcast the ID information of the person and the wearable device information corresponding to the ID information to urge the relevant staff to wear the corresponding wearable device in time. In addition, the corresponding ID information and violation can be sent to the terminal device for recording and storage.

[0083] See also Figure 2 , which is a schematic diagram of an image-based wearable device recognition system provided in an embodiment of the present invention; as shown in the figure, the system 200 in this embodiment may include: an acquisition unit 201, a first judgment unit 202, a face recognition unit 203, a first determination unit 204, a second judgment unit 205, a second determination unit 206, a third judgment unit 207, and an alarm unit 208. Among them, the acquisition unit 201 is used to acquire the detection image; the first judgment unit 202 is used to judge whether there is a human body in the detection image; the face recognition unit 203 is used to perform face recognition on the human body to obtain the ID information corresponding to the human body; the first determination unit 204 is used to determine the wearable device that the human body needs to wear according to the human body ID information; the second judgment unit 205 is used to judge whether there is a wearable device that the human body needs to wear in the detection image; the second determination unit 206, if the wearable device that needs to be worn exists in the detection image, obtains the posture of the human body, and determines the correct wearing area according to the posture of the human body; the third judgment unit 207 is used to judge whether the wearable device that needs to be worn is in the correct wearing area; the alarm unit 208 is used to determine whether the wearable device that needs to be worn is correctly worn by the human body if the wearable device that needs to be worn is in the correct wearing area, and performs an alarm operation if the wearable device that needs to be worn is not in the correct wearing area.

[0084] In one embodiment, the acquisition unit 201 acquires real-time video streams transmitted from multiple cameras, and parses the real-time video streams to obtain a picture to be analyzed; obtains the optical flow component value of the picture to be analyzed; if the optical flow component value of the picture to be analyzed is greater than a first preset threshold, the picture to be analyzed is selected as a detection picture.

[0085] In one embodiment, the system 200 also includes a model training unit 209, which is used to collect a variety of wearable device image samples and non-wearable device image samples, classify, label and train them, and obtain a wearable device feature model; collect human image samples and non-human image samples, classify, label and train them, and obtain a human feature model.

[0086] In one embodiment, the second determining unit 206 is specifically used to obtain the aspect ratio α of the human body frame; if the value of α is greater than the second preset threshold, it is determined that the posture of the human body is standing; if the value of α is less than the second preset threshold, it is determined that the posture of the human body is squatting; if the posture of the human body is standing, the correct wearing area is determined to be the first helmet wearing area and / or the first safety rope wearing area and / or the first glove wearing area, the first helmet wearing area is a square area greater than the first preset height on the upper part of the human body frame; the first safety rope wearing area is a square area from the second preset height to the first preset height in the middle of the human body frame, and the first glove wearing area is a square area from the third preset height to the second preset height in the lower part of the human body frame; if the posture of the human body is squatting, the correct wearing area is determined to be the second helmet wearing area and / or the second safety rope wearing area and / or the second glove wearing area, the second helmet wearing area is a square area greater than the fourth preset height on the upper part of the human body frame; the second safety rope wearing area is a square area from the fifth preset height to the fourth preset height in the middle of the human body frame, and the second glove wearing area is a square area below the fifth preset height in the lower part of the human body frame.

[0087] In one embodiment, the first judgment unit 202 is specifically used to nominate using k different rectangular frames on the convolutional feature layer of the detection image to screen out candidate areas where human bodies may exist, where k is a positive integer; perform human body feature extraction on the candidate areas respectively; classify and regress the features using a human body feature model to identify whether there is a human body in the detection image; if it is identified that there is a human body in the detection image, perform border position regression processing on the human body to obtain a human body frame.

[0088] In one embodiment, the second judgment unit 205 is specifically used to nominate using k different rectangular boxes on the convolutional feature layer of the detection image to screen out candidate areas where wearable devices may exist, where k is a positive integer; extract features of wearable devices from the candidate areas respectively; classify and regress the features using multiple wearable device feature models to identify whether the detection image contains a wearable device that needs to be worn by a human body; if it is identified that the detection image contains a wearable device that needs to be worn by a human body, perform border position regression processing on the wearable device that needs to be worn by the human body to obtain a wearable device frame.

[0089] In one embodiment, the third judgment unit 207 is used to obtain a wearable device that has an overlapping area with the human body frame of the human body and the overlapping area is in the head, torso, and hand of the human body as a target wearable device; if the posture of the human body is standing, it is judged whether the wearable device frame of the target wearable device is respectively located in the first helmet wearing area, the first safety rope wearing area, and the first glove wearing area; if so, it is judged that the target wearable device is in the correct wearing area; if the posture of the human body is squatting, it is judged whether the wearable device frame of the target wearable device is respectively located in the second helmet wearing area, the second safety rope wearing area, and the second glove wearing area; if so, it is judged that the target wearable device is in the correct wearing area. The above disclosed content is only a preferred feasible embodiment of the present invention, and does not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated with the development of technology.

Claims

1. An image-based wearable device recognition system, characterized in that: include: An acquisition unit, used for acquiring a detection image; A first judging unit, used to judge whether there is a human body in the detection picture; A face recognition unit, used to perform face recognition on a human body to obtain ID information corresponding to the human body; a first determination unit, used to determine a wearable device that the human body needs to wear according to the human body ID information; The second judgment unit is used to judge whether there is a wearable device that needs to be worn by a human body in the detection picture; The second determination unit is used to obtain the posture of the human body if the wearable device to be worn exists in the detection image, and determine the correct wearing area according to the posture of the human body; the third judgment unit is used to judge whether the wearable device to be worn is in the correct wearing area; the alarm unit is used to determine whether the human body is correctly wearing the wearable device to be worn if the wearable device to be worn is in the correct wearing area, and to perform an alarm operation if the wearable device to be worn is not in the correct wearing area.

2. A wearable device recognition method based on an image, characterized in that: The following steps are involved: S1, obtain the detection picture; S2, determining whether there is a human body in the detection image; S3, if there is a human body in the detection image, perform face recognition on the human body to obtain ID information corresponding to the human body, and obtain a wearable device that the human body needs to wear according to the ID information; S4, determining whether the wearable device to be worn exists in the detection image; S5, if the detection image contains the wearable device to be worn, obtaining the posture of the human body, and determining the correct wearing area according to the posture of the human body; S6, determining whether the wearable device to be worn is in the correct wearing area; S7: If yes, determine whether the wearable device is worn correctly by the human body; if no, perform an alarm operation.

3. The image-based wearable device recognition method according to claim 2, characterized in that: The step S1 includes: acquiring a real-time video stream through a camera, parsing the real-time video stream to obtain a picture to be analyzed, the picture to be analyzed includes an intra-frame coded frame and a predictive coded frame, and obtaining an optical flow component value of the picture to be analyzed; if the optical flow component value of the picture to be analyzed is greater than a first preset threshold, selecting the picture to be analyzed as a detection picture.

4. The image-based wearable device recognition method according to claim 3, characterized in that: The step S2 includes: collecting human image samples and non-human image samples, classifying, labeling and training them, and obtaining a human feature model; using k different rectangular frames to nominate on the convolution feature layer of the detection image, and screening out candidate areas where human bodies may exist, where k is a positive integer; extracting human body features from the candidate areas respectively; using the human feature model to classify and regress the features, and identify whether the detection image contains a human body; if it is identified that the detection image contains a human body, performing border position regression processing on the human body to obtain a human body frame.

5. The image-based wearable device recognition method according to claim 4, characterized in that: The step S3 includes: storing the correspondence between the ID information of the human body and the wearable device to be worn, and when the ID information of the human body is identified, directly obtaining one or more wearable devices to be worn by the human body according to the correspondence.

6. The image-based wearable device recognition method according to claim 5, characterized in that: The step S4 includes: collecting a variety of wearable device image samples and non-wearable device image samples, classifying, labeling and training, and obtaining a variety of wearable device models; using k different rectangular frames to nominate on the convolutional feature layer of the detection image, and screening out candidate areas where wearable devices may exist, where k is a positive integer; extracting features of wearable devices from the candidate areas respectively; using the wearable device model to classify and regress the features, and identify whether the detection image has a wearable device to be worn; if it is identified that the detection image has a wearable device to be worn, performing border position regression processing on the wearable device to be worn, and obtaining a wearable device frame.

7. The image-based wearable device recognition method according to claim 6, characterized in that: The acquiring of the posture of the human body comprises: acquiring the aspect ratio α of the human body frame; if the value of α is greater than a second preset threshold, determining that the posture of the human body is standing; if the value of α is less than the second preset threshold, determining that the posture of the human body is squatting.

8. The image-based wearable device identification method according to claim 7, characterized in that: The wearable device includes a helmet, a safety rope and gloves, and the determining of the correct wearing area according to the posture of the human body includes: if the posture of the human body is standing, determining the correct wearing area as a first helmet wearing area and / or a first safety rope wearing area and / or a first glove wearing area according to the ID information, the first helmet wearing area being a square area greater than a first preset height at the upper part of the human body frame; the first safety rope wearing area being a square area from a second preset height to a first preset height in the middle part of the human body frame, and the first glove wearing area being a square area from a third preset height to a second preset height in the lower part of the human body frame; if the posture of the human body is squatting, determining the correct wearing area as a second helmet wearing area and / or a second safety rope wearing area and / or a second glove wearing area according to the ID information, the second helmet wearing area being a square area greater than a fourth preset height at the upper part of the human body frame; the second safety rope wearing area being a square area from a fifth preset height to a fourth preset height in the middle part of the human body frame, and the second glove wearing area being a square area below the fifth preset height in the lower part of the human body frame.

9. The image-based wearable device recognition method according to claim 8, characterized in that: Obtain a wearable device whose overlapping area with the human body frame and the overlapping area on the head, torso, and hand of the human body as the target wearable device; if the posture of the human body is standing, determine whether the wearable device frame of the target wearable device is respectively located in the first safety helmet wearing area, the first safety rope wearing area, and the first glove wearing area; if so, determine that the target wearable device is in the correct wearing area; if the posture of the human body is squatting, determine whether the wearable device frame of the target wearable device is respectively located in the second safety helmet wearing area, the second safety rope wearing area, and the second glove wearing area; if so, determine that the target wearable device is in the correct wearing area.

10. The image-based wearable device recognition method according to claim 9, characterized in that: In the step S7: when it is detected that a human body is not wearing the wearable device correctly, the ID information of the human body and the wearable device information corresponding to the ID information are broadcasted.

Images