CN110135476A - A kind of detection method of personal safety equipment, device, equipment and system - Google Patents
A kind of detection method of personal safety equipment, device, equipment and system Download PDFInfo
- Publication number
- CN110135476A CN110135476A CN201910347733.5A CN201910347733A CN110135476A CN 110135476 A CN110135476 A CN 110135476A CN 201910347733 A CN201910347733 A CN 201910347733A CN 110135476 A CN110135476 A CN 110135476A
- Authority
- CN
- China
- Prior art keywords
- personal safety
- safety equipment
- detection result
- target
- detection
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Abstract
The present embodiments relate to image procossings and machine learning techniques field, disclose detection method, device, equipment and the system of a kind of personal safety equipment.The wherein detection method of the personal safety equipment is applied to intelligent terminal, which comprises training personal safety equipment's detection model in advance；Obtain the video flowing of construction site；Each frame image in the video flowing is pre-processed；Pretreated each frame image is inputted into personal safety equipment's detection model, generates Preliminary detection result；The Preliminary detection result is post-processed, final detection result is generated.Through the above way, the embodiment of the present invention solves the low technical problem of Detection accuracy existing for conventional security cap, safety vest detection method, automatic monitoring is realized without the possibility risk situation of personal safety equipment to monitoring center and early warning risk, and improves Detection accuracy.
Description
Technical field
The present invention relates to image procossings and machine learning techniques field, more particularly to the detection of personal safety equipment a kind of
Method, apparatus, equipment and system.
Background technique
Safety cap and safety vest are all trades and professions safety in production person and the essential safety of high altitude operation personnel
Apparatus, accurate measurements constructor whether safe wearing cap and safety vest it is extremely important.With artificial intelligence technology in recent years
It continues to develop, traditional safety management constantly develops to wisdom safety management, reduces human resources, in real time automation pipe to reach
Reason, the in time targets such as alarm.
Conventional security cap, safety vest detection method are based primarily upon image procossing and machine learning method to judge a people
Whether safe wearing cap and safe wearing vest.Adaptability is poor usually under complex environment for such methods, and there are more
Erroneous detection and missing inspection, Detection accuracy it is low.
Summary of the invention
The embodiment of the present invention is intended to provide detection method, device, equipment and the system of a kind of personal safety equipment, solves
The low technical problem of Detection accuracy existing for conventional security cap, safety vest detection method, improves personal safety equipment's
Detection accuracy.
In order to solve the above technical problems, the embodiment of the present invention the following technical schemes are provided:
In a first aspect, the embodiment of the present invention provides the detection method of personal safety equipment a kind of, which comprises
Training personal safety equipment's detection model in advance；
Obtain the video flowing of construction site；
Each frame image in the video flowing is pre-processed；
Pretreated each frame image is inputted into personal safety equipment's detection model, generates Preliminary detection result；
The Preliminary detection result is post-processed, final detection result is generated.
In embodiments of the present invention, the method also includes:
According to the final detection result, the work of non-safe wearing cap and/or non-safe wearing vest is judged whether there is
Make personnel, if it exists the staff of non-safe wearing cap and/or non-safe wearing vest, then it is assumed that the construction site has peace
Full blast danger, reports monitoring center for the security risk.
In embodiments of the present invention, preparatory trained personal safety equipment's detection model, specifically includes:
Collect the multiple image of the construction site；
Target in each described image is labeled, markup information is generated；
The image that mark is completed is trained, personal safety equipment's detection model is generated.
In embodiments of the present invention, personal safety equipment's detection model is feedforward convolutional neural networks, the feedforward
Convolutional neural networks include: base net network and a series of convolution kernels, and the base net network is used for each frame figure in the video flowing
As carrying out feature extraction, the feature that convolution kernel is used to be extracted according to the base net network carries out target detection, to generate the target
Classification and the target position.
In embodiments of the present invention, personal safety equipment's detection model includes training loss function, the training damage
Lose the Sigmoid Classification Loss function and weight smooth L that function is weighting1The sum of loss function is positioned, the weighting
Sigmoid Classification Loss function classifies to the target in described image for optimizing personal safety equipment's model, institute
It states and weights smooth L1Positioning loss function is for optimizing the position that personal safety equipment's model determines the target.
In embodiments of the present invention, each frame image in the video flowing pre-processes, comprising:
Described image is zoomed in and out into preset resolution ratio.
In embodiments of the present invention, described that the Preliminary detection result is post-processed, generate final detection result, tool
Body includes:
The Preliminary detection result is screened by non-maxima suppression algorithm, generates final detection result.
Second aspect, the embodiment of the present invention provide the detection device of personal safety equipment a kind of, set applied to intelligent terminal
Standby, described device includes:
Personal safety equipment's detection model unit, for training personal safety equipment's detection model in advance；
Video streaming units, for obtaining the video flowing of construction site；
Pretreatment unit, for being pre-processed to each frame image in the video flowing；
Preliminary detection result unit detects mould for pretreated each frame image to be inputted the personal safety equipment
Type generates Preliminary detection result；
Final detection result unit generates final detection result for post-processing to the Preliminary detection result；
Alarm Unit, for judging whether there is non-safe wearing cap and/or not wearing according to the final detection result
The staff of safety vest, the if it exists staff of non-safe wearing cap and/or non-safe wearing vest, then it is assumed that this is applied
There are security risks at work scene, and the security risk is reported monitoring center.
In embodiments of the present invention, personal safety equipment's detection model unit, is specifically used for:
Collect the multiple image of the construction site；
Target in each described image is labeled, markup information is generated；
The image that mark is completed is trained, personal safety equipment's detection model is generated.
In embodiments of the present invention, personal safety equipment's detection model is feedforward convolutional neural networks, the feedforward
Convolutional neural networks include: base net network and a series of convolution kernels, and the base net network is used for each frame figure in the video flowing
As carrying out feature extraction, the feature that convolution kernel is used to be extracted according to the base net network carries out target detection, to generate the target
Classification and the target position.
In embodiments of the present invention, personal safety equipment's detection model includes training loss function, the training damage
Lose the Sigmoid Classification Loss function and weight smooth L that function is weighting1The sum of loss function is positioned, the weighting
Sigmoid Classification Loss function classifies to the target in described image for optimizing personal safety equipment's model, institute
It states and weights smooth L1Positioning loss function is for optimizing the position that personal safety equipment's model determines the target.
In embodiments of the present invention, the pretreatment unit, is specifically used for:
Described image is zoomed in and out into preset resolution ratio.
In embodiments of the present invention, the final detection result unit, is specifically used for:
The testing result is screened by non-maxima suppression algorithm, obtains final detection result.
The third aspect, the embodiment of the present invention provide a kind of intelligent terminal, comprising:
At least one processor；And
The memory being connect at least one described processor communication；Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
A processor executes, so that at least one described processor is able to carry out the detection method of personal safety equipment as described above.
Fourth aspect, the embodiment of the present invention provide a kind of safety monitoring system, the system comprises:
Above-mentioned intelligent terminal, the intelligent terminal integrate photographic device, locality connection peripheral hardware camera or
Pass through the IP Camera of internet communication；
The monitoring center being connect with the intelligent terminal by internet；And
Receive the user terminal of the monitoring center push security risk.
5th aspect, the embodiment of the invention also provides a kind of non-volatile computer readable storage medium storing program for executing, the calculating
Machine readable storage medium storing program for executing is stored with computer executable instructions, and the computer executable instructions are for enabling intelligent terminal
Enough execute the detection method of personal safety equipment as described above.
The beneficial effect of the embodiment of the present invention is: being in contrast to the prior art down, provided in an embodiment of the present invention one
The detection method of kind personal safety equipment, passes through and trains personal safety equipment's detection model in advance；Obtain the video of construction site
Stream；Each frame image in the video flowing is pre-processed；By pretreated each frame image input personal peace
Full equipment detection model, generates Preliminary detection result；The Preliminary detection result is post-processed, final detection knot is generated
Fruit.By the above-mentioned means, the embodiment of the present invention solves Detection accuracy existing for conventional security cap, safety vest detection method
Low technical problem realizes automatic monitoring without the possibility risk situation of personal safety equipment to monitoring center and pre- police conduct
Danger, and improve Detection accuracy.
Detailed description of the invention
One or more embodiments are illustrated by the picture in corresponding attached drawing, these exemplary theorys
The bright restriction not constituted to embodiment, the element in attached drawing with same reference numbers label are expressed as similar element, remove
Non- to have special statement, composition does not limit the figure in attached drawing.
Fig. 1 is the system schematic of the first application environment provided in an embodiment of the present invention；
Fig. 2 is the system schematic of second of application environment provided in an embodiment of the present invention；
Fig. 3 is a kind of flow diagram of the detection method of personal safety equipment provided in an embodiment of the present invention；
Fig. 4 is the refined flow chart of the step S31 in Fig. 3；
Fig. 5 is a kind of schematic diagram of Preliminary detection result provided in an embodiment of the present invention；
Fig. 6 is a kind of schematic diagram of final detection result provided in an embodiment of the present invention；
Fig. 7 is a kind of schematic diagram of the workflow of the detection method of personal safety equipment provided in an embodiment of the present invention；
Fig. 8 is a kind of schematic diagram of the detection device of personal safety equipment provided in an embodiment of the present invention；
Fig. 9 is a kind of structural schematic diagram of intelligent terminal provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
In addition, as long as technical characteristic involved in the various embodiments of the present invention described below is each other not
Constituting conflict can be combined with each other.
Supervised learning, refers to through existing training sample (i.e. given data and its corresponding output) Lai Xunlian, from
And an optimal models are obtained, recycle this model to be mapped as exporting as a result, to defeated accordingly by all new data samples
Result carries out simple judgement to realize the purpose of classification out, then this optimal models be also just provided with to unknown data into
The ability of row classification.
It rolls up root neural network (convolutional neural network), refers at least in a certain layer kind convolution
Operation (convolution) replaces the neural network of matrix multiplication.The characteristic of convolution algorithm determines that neural network is suitable for
Handle the data with fenestral fabric.Most typical grid type data are exactly digital picture, either gray level image or coloured silk
Chromatic graph picture, one group of scalar or the vector being all defined on two-dimensional pixel grid.Thus convolutional neural networks are since birth, just
It is widely used among image and text identification, and gradually expands to the other fields such as natural language processing.
Convolution is a kind of mathematical operation carried out to two functions, there is different interpretive modes in different subjects.?
In convolutional network, the function of two participation operations is called input and kernel function (kernel function) respectively.Essentially,
Convolution is exactly that the process of summation is weighted to input using kernel function as weight coefficient.
Image to be processed is converted one or more picture element matrix by input layer, and convolutional layer utilizes one or more volumes
Product core extracts feature from picture element matrix, and obtained Feature Mapping is admitted to pond layer after nonlinear function is handled, by pond
Change layer and executes dimensionality reduction operation.Convolutional layer and being used alternatingly for pond layer can be such that convolutional neural networks extract in different levels
Characteristics of image.Input of the feature finally obtained as full articulamentum, by the classifier output category result of full articulamentum.
In the training of convolutional neural networks, parameter to be trained is the weight coefficient in convolution kernel, that is, convolutional layer
Matrix.What training used is also the method for backpropagation, and the continuous renewal of parameter is able to ascend the precision of image characteristics extraction.
In embodiments of the present invention, personal safety equipment's detection model is SSD detection model, personal safety equipment's inspection
Surveying model is feedforward convolutional neural networks, and SSD is the abbreviation of Single Shot MultiBox Detector, and SSD algorithm is one
Kind directly predicts the classification of target and the multi-target detection algorithm of position.
Referring to Fig. 1, Fig. 1 is the system schematic of the first application environment provided in an embodiment of the present invention；
As shown in Figure 1, the system includes: IP Camera, server, monitoring center and user terminal, wherein described
IP Camera wireless communication connection server, such as: the server passes through bluetooth, WLAN, 2G, 3G, 4G, 5G etc.
Mode connects IP Camera, and the server connects monitoring center, and the server (is not dressed for obtaining security risk
Personal safety equipment), and report security risk to monitoring center, the monitoring center connects user terminal, is used for the peace
Full blast nearly pushes to user terminal, and the user terminal is the terminal device of monitoring area safety director.
Referring to Fig. 2, Fig. 2 is the system schematic of second of application environment provided in an embodiment of the present invention；
As shown in Fig. 2, the system comprises: monitoring center, intelligent terminal and user terminal, the system are applicable in
In the environment of narrowband Internet of Things definite transmission ability, wherein the intelligent terminal passes through local serial ports connection, optical cable connection
Local camera or the integrated local camera of the intelligent terminal；Wherein, the intelligent terminal passes through Internet of Things
Security risk, such as NB-lot are reported, the user terminal is attached with the monitoring center by wireless network, such as: institute
It states user terminal to communicate to connect by the modes such as bluetooth, WLAN, 2G, 3G, 4G, 5G and the monitoring center, the use
Family terminal is used to receive the warning message or security risk information that the monitoring center is sent.
Wherein, security risk early warning can be pushed to the safety director movement of the construction site eventually by the monitoring center
End or PC terminal.
In an embodiment of the present invention, intelligent terminal can be camera terminal with computing capability, personal meter
The energy such as calculation machine, smart phone, palm PC (Personal Digital Assistant, PDA), tablet computer, smartwatch
The electronic equipment of display alarm information.
It specifically, with intelligent terminal is below the camera terminal with computing capability, user terminal is intelligent hand
The embodiment of the present invention is specifically addressed for machine.
Referring to Fig. 3, Fig. 3 is a kind of process signal of the detection method of personal safety equipment provided in an embodiment of the present invention
Figure；
As shown in figure 3, the method is applied to intelligent terminal, for example, the camera terminal with computing capability, institute
The method of stating includes:
Step S31: training personal safety equipment's detection model in advance；
Specifically, the personal safety equipment includes: the personal safety equipments such as safety cap, safety vest, Labor protection shoes, it is described
Personal safety equipment's detection model is SSD detection model, in order to enable SSD detection model is able to detect safety cap and safety back
The heart is needed to be trained the SSD detection model in advance using a large amount of image, then can just be deployed to true application scenarios
In.
In embodiments of the present invention, personal safety equipment's detection model is the feedforward convolutional neural networks of open source
(Single Shot MultiBox Detector,SSD).Convolutional neural networks feedover mainly by base net network and a series of convolution
Core composition, wherein the base net network is MobilenetV1 neural network, MobilenetV1 neural network is one and is suitable for moving
The lightweight convolutional neural networks of moved end, it is designed for classifying.In SSD detection model, the MobilenetV1 mind
It is intercepted classification layer through network, the MobilenetV1 neural network for being intercepted classification layer is taken as feature extractor,
For extracting the feature of image, wherein the feature of extraction is input to a series of convolution kernels in SSD for target detection, often
One convolution kernel is equivalent to a classifier, each convolution kernel is responsible for the classification of one target of detection and the position of target.?
In the safety cap of this system, safety vest test problems, 4 classifications are predominantly detected: 1. safe wearing cap (withHelmet), 2.
Non- safe wearing cap (withoutHelmet), 3. safe wearing vest (withVest), 4. non-safe wearing vest
(withoutVest), that is, need to judge someone whether safe wearing cap and safety vest.
In embodiments of the present invention, when to SSD detection model input picture, the SSD detection model needs to detect institute
It states and whether there is the target for belonging to 4 classifications specified above in image, and export target and belong to certain a kind of confidence
With the position of target.If a possibility that confidence is higher, and target belongs to corresponding classification is bigger；The position of target
It is usually provided by two coordinate points, the two coordinate points have determined that the minimum rectangle for surrounding the target surrounds frame.
It is the refined flow chart of the step S31 in Fig. 3 referring again to Fig. 4, Fig. 4；
As shown in figure 4, preparatory trained personal safety equipment's detection model, comprising:
Step S311: the multiple image of the construction site is collected；
Specifically, include the multiple image of the construction site of people from interconnection online collection, in described image comprising wearing or
The people of non-safe wearing cap and wearing or non-safe wearing vest, such as: parts of images only includes and wears in the multiple image
The people to wear a safety helmet, parts of images include the people of non-safe wearing cap, and parts of images includes the people of safe wearing vest, part figure
People of the picture comprising non-safe wearing vest covers all possible situation by the multiple image of the collection construction site, when
When the quantity of image is enough, personal safety equipment's detection model can be preferably trained.
Step S312: being labeled the target in each described image, generates markup information；
Specifically, carrying out target detection (object detection) to image, wherein the target in image refers to figure
The interested object as present in, such as head and the upper body of people.
After described image is collected, needing further to be labeled every image could be used to that SSD to be trained to examine
Model is surveyed, because the process of training SSD is a supervised learning process, mark is necessary.Emerging is felt to each of image
Interesting target is labeled, and generates markup information, and for each target, markup information is the classification and target of this target
Position, the classification of the target include: safe wearing cap (withHelmet), non-safe wearing cap, safe wearing vest and
Non- safe wearing vest can be used for training SSD model after all image labelings.
Wherein, since SSD is that a general detection model is needed when being applied to different concrete scenes according to application
The demand of scene is appropriately modified SSD.Since this system needs to detect 4 classifications, therefore the output classification of SSD is modified to 4.
Wherein, the position of the target refers to the position of target in the picture.It usually will with a minimum circumscribed rectangle
Target frames, and the position of this rectangle is exactly the position of target, and the position of rectangle is usually by the upper left corner of rectangle and the lower right corner
Coordinate determines.
Step S313: the image that mark is completed is trained, and generates personal safety equipment's detection model；
Specifically, personal safety equipment's detection model, i.e. SSD detection model, including training loss function, the instruction
Practice the Sigmoid Classification Loss function and weight smooth L that loss function includes weighting1Position loss function, the training damage
Lose the Sigmoid Classification Loss function and weight smooth L that function L is weighting1Position the sum of loss function, wherein described to add
The Sigmoid Classification Loss function of power divides the target in described image for optimizing personal safety equipment's model
Class, the smooth L of the weighting1Positioning loss function is for optimizing the position that personal safety equipment's model determines the target
It sets.
In embodiments of the present invention, due to having used the Sigmoid Classification Loss function (Weighted weighted simultaneously
Sigmoid Classification Loss) and the smooth L of weighting1Position loss function (Weighted Smooth
L1Localization Loss) as total loss function training come to train SSD, total loss function L be two loss functions
Addition.Wherein, the effect of the Sigmoid Classification Loss function of weighting is mainly used for target classification, is judgement target
Whether safe wearing cap, whether safe wearing vest；Weight smooth L1The effect of positioning loss function is mainly used for returning
The coordinate position of target in the picture out can effectively train personal safety equipment's detection model.
In embodiments of the present invention, the method also includes: when personal safety equipment's model, i.e., the described SSD detection
After the completion of model configuration, using RMSprop optimizer with the learning rate damped manner training SSD detection model, and in training
The quantity of described image is expanded in such a way that random cropping and Random Level are overturn in the process.
Specifically, training SSD was easy to intend on fewer image since the parameter of SSD detection model has up to ten thousand
It closes, causes the Generalization Ability for training the SSD come poor.In order to reduce over-fitting risk, generally can all select to increase image
Quantity, random cropping and Random Level overturning are two kinds of common image amplification modes.Wherein, the random cropping refers to
Inside one image, random cropping goes out multiple small images, and the flip horizontal refers to image to be spun upside down.By with
Machine is cut and Random Level overturning amplification image, thus increase amount of images, reduction over-fitting risk.
In embodiments of the present invention, the target detection platform (Tensorflow that entire training process is increased income in Google
Object Detection API) on complete.
Step S32: the video flowing of construction site is obtained；
Specifically, the construction site is equipped with multiple cameras, the construction site is obtained by the camera
Video flowing, the video flowing are made of multiple image.
Step S33: each frame image in the video flowing is pre-processed；
Specifically, each frame image in the video flowing pre-processes, comprising: described image contracts
It is put into preset resolution ratio.Due to the influence of the calculation amount and complexity of SSD detection model, also, in order not to lose each frame
The information of image, it is therefore desirable to each frame image in the video flowing is pre-processed, such as: described image is contracted
It is put into preset resolution ratio, such as: the preset resolution ratio is 300*300, it is to be understood that can also be according to specific
Need to adjust the preset resolution ratio, such as: 500*500 is set by the preset resolution ratio, and etc..
Step S34: pretreated each frame image is inputted into personal safety equipment's detection model, generates preliminary inspection
Survey result；
Specifically, will be examined by the personal safety equipment after the completion of personal safety equipment's detection model training
It surveys model to detect the image of the video flowing of the construction site, generates Preliminary detection result.The Preliminary detection result
It include: the position of the classification of target, the corresponding confidence of the classification of target and target.Wherein, the Preliminary detection knot
Fruit includes the position of the classification of multiple targets, the corresponding confidence of the classification of target and target.
Wherein, when the corresponding each frame of the video flowing for inputting the construction site to personal safety equipment's detection model
When image, personal safety equipment's detection model, which needs to detect to whether there is in described image, belongs to 4 classes specified above
Other target, and export the position that target belongs to certain a kind of confidence and target.If the confidence is higher,
It is bigger that target belongs to a possibility that corresponding classification；The position of target is usually provided by two coordinate points, the two coordinate points are true
The minimum rectangle for having determined to surround the target surrounds frame.
Referring to Fig. 5, Fig. 5 is a kind of schematic diagram of Preliminary detection result provided in an embodiment of the present invention；
As shown in figure 5, generating Preliminary detection as a result, institute after the crash helmet security detection model detects image
The multiple rectangles encirclement frames of generation may be detected by stating a target in Preliminary detection result, and each rectangle surrounds the corresponding target of frame
Classification, the corresponding confidence of the classification of target and target position.In order to more accurately determine the target category
And target position, it needs to post-process the Preliminary detection result.
Step S35: post-processing the Preliminary detection result, generates final detection result.
Specifically, described post-process the Preliminary detection result, final detection result is generated, is specifically included: is logical
It crosses non-maxima suppression algorithm to screen the Preliminary detection result, generates final detection result.Wherein, described non-very big
It is worth restrainable algorithms (non-maximum suppression, NMS), for the preliminary of personal safety equipment's detection model
Testing result is further screened, and the testing result of redundancy is removed, and generate final detection result.Wherein, described final
Testing result includes: the position of the classification of target, the corresponding confidence of the classification of target and target.It is understood that
It is that the corresponding final detection result is a testing result, and the testing result includes the classification of the target, target
The position of the corresponding confidence of classification and target.
Referring to Fig. 6, Fig. 6 is a kind of schematic diagram of final detection result provided in an embodiment of the present invention；
As shown in fig. 6, screening by non-maxima suppression algorithm to the Preliminary detection result, final detection is generated
As a result, the final detection result is the classification of target and the position of target, wherein the classification of the target includes: to wear
The position of safety cap, non-safe wearing cap, safe wearing vest and non-safe wearing vest, the target passes through two coordinates
Point determines, and generates rectangle according to described two coordinate points and surround frame, wherein described two coordinate points are respectively the rectangle packet
The upper left angle point and bottom right angle point of peripheral frame.As shown in fig. 6, detecting two targets, the classification of the target, which is respectively as follows:, not to be worn
Safety cap and safe wearing vest, the confidence of the non-safe wearing cap are 92%, and the safe wearing vest is set
Confidence score is 99%.
It is a kind of workflow of the detection method of personal safety equipment provided in an embodiment of the present invention referring again to Fig. 7, Fig. 7
The schematic diagram of journey；
As shown in fig. 7, the workflow of the detection method of the personal safety equipment, comprising:
Step S71: the video flowing of construction site is acquired；
Specifically, the construction site is equipped with multiple cameras, the construction site is obtained by the camera
Video flowing, the video flowing are made of multiple image.
Step S72: each frame image in the video flowing is pre-processed；
Specifically, each frame image in the video flowing pre-processes, comprising: described image contracts
It is put into preset resolution ratio.Due to the influence of the calculation amount and complexity of SSD detection model, also, in order not to lose each frame
The information of image, it is therefore desirable to each frame image in the video flowing is pre-processed, such as: described image is contracted
It is put into preset resolution ratio, such as: the preset resolution ratio is 300*300, it is to be understood that can also be according to specific
Need to adjust the preset resolution ratio, such as: 500*500 is set by the preset resolution ratio, and etc..
Step S73: pretreated each frame image is inputted into personal safety equipment's detection model, generates preliminary inspection
Survey result；
Specifically, will be examined by the personal safety equipment after the completion of personal safety equipment's detection model training
It surveys model to detect the image of the video flowing of the construction site, generates Preliminary detection result.The Preliminary detection result
It include: the position of the classification of target, the corresponding confidence of the classification of target and target.Wherein, the Preliminary detection knot
Fruit includes the position of the classification of multiple targets, the corresponding confidence of the classification of target and target.
Wherein, when the corresponding each frame of the video flowing for inputting the construction site to personal safety equipment's detection model
When image, personal safety equipment's detection model, which needs to detect to whether there is in described image, belongs to 4 classes specified above
Other target, and export the position that target belongs to certain a kind of confidence and target.If the confidence is higher,
It is bigger that target belongs to a possibility that corresponding classification；The position of target is usually provided by two coordinate points, the two coordinate points are true
The minimum rectangle for having determined to surround the target surrounds frame.
Step S74: post-processing the detection model, generates final detection result；
Specifically, described post-process the Preliminary detection result, final detection result is generated, is specifically included: is logical
It crosses non-maxima suppression algorithm to screen the Preliminary detection result, generates final detection result.Wherein, described non-very big
It is worth restrainable algorithms (non-maximum suppression, NMS), for the preliminary of personal safety equipment's detection model
Testing result is further screened, and the testing result of redundancy is removed, and generate final detection result.Wherein, described final
Testing result includes: the position of the classification of target, the corresponding confidence of the classification of target and target.It is understood that
It is that the corresponding final detection result is a testing result, and the testing result includes the classification of the target, target
The position of the corresponding confidence of classification and target.
Step S75: judge staff whether safe wearing cap and safety vest；
Specifically, the final detection result includes the classification of target, the classification of target according to the final detection result
The position of corresponding confidence and target, according to the corresponding confidence of the classification of the classification of the target and the target
Spend score, judge staff whether safe wearing cap and safety vest, if the non-safe wearing cap of the staff or not wearing
Safety vest is worn, then enters step S76: triggering alarm.
In embodiments of the present invention, described that pretreated each frame image is inputted into personal safety equipment's detection mould
After type, if there are interesting targets in image, it will 4 confidences of output respectively correspond 4 target categories: 1. wearing peace
Full cap (withHelmet), 2. non-safe wearing cap (withoutHelmet), 3. safe wearing vest (withVest), 4. not
Safe wearing vest (withoutVest).It is by obtaining 4 confidences, the confidence highest is corresponding
Classification, be determined as the target category that detected.Such as: 4 confidences of output are respectively as follows: 0.05,0.1,0.8,
0.05；Corresponding classification sequence are as follows: 1. safe wearing cap (withHelmet), 2. non-safe wearing cap (withoutHelmet),
3. safe wearing vest (withVest), 4. non-safe wearing vest (withoutVest).It is highest in 4 confidences
It is 0.8, and it is 3. safe wearing vest (withVest) that highest confidence is corresponding, it is determined that the target category detected
In 3. safe wearing vest (withVest).
Step S76: triggering alarm；
In embodiments of the present invention, if the intelligent terminal detects the non-safe wearing cap of the staff or not
Safe wearing vest, then the intelligent terminal will report security risk, alternatively, intelligent terminal connection alarm dress
It sets, and sends alarm command to the alarm device, so that the alarm device is alarmed, the alarm side of the alarm device
Formula include voice reminder, text remind, and etc..In embodiments of the present invention, the intelligent terminal is also connected with intelligence
Terminal, the intelligent terminal are the portable intelligent terminal of the staff, the triggering alarm, comprising: Xiang Suoshu work
The intelligent terminal for making personnel sends alarm command, so that the intelligent terminal of the staff reminds the staff to wear peace
Full cap and/or safety vest.
In embodiments of the present invention, it by providing the detection method of personal safety equipment a kind of, is set applied to intelligent terminal
It is standby, which comprises training personal safety equipment's detection model in advance；Obtain the video flowing of construction site；To the video
Each frame image in stream is pre-processed；Pretreated each frame image is inputted into the personal safety equipment and detects mould
Type generates Preliminary detection result；The Preliminary detection result is post-processed, final detection result is generated.Pass through above-mentioned side
Formula, the embodiment of the present invention solve the low technical problem of Detection accuracy existing for conventional security cap, safety vest detection method,
Automatic monitoring is realized without the possibility risk situation of personal safety equipment to monitoring center and early warning risk, and it is quasi- to improve detection
True rate.
Referring to Fig. 8, Fig. 8 is a kind of schematic diagram of the detection device of personal safety equipment provided in an embodiment of the present invention；
The detection device of the personal safety equipment can be applied to intelligent terminal, such as: Cloud Server, Intelligent terminal for Internet of things are set
It is standby.
As shown in figure 8, the detection device 80 of the personal safety equipment includes:
Personal safety equipment's detection model unit 81, for training personal safety equipment's detection model in advance；
Video streaming units 82, for obtaining the video flowing of construction site；
Pretreatment unit 83, for being pre-processed to each frame image in the video flowing；
Preliminary detection result unit 84 is detected for pretreated each frame image to be inputted the personal safety equipment
Model generates Preliminary detection result；
Final detection result unit 85 generates final detection result for post-processing to the Preliminary detection result；
Alarm Unit 86, for judging whether there is non-safe wearing cap and/or not wearing according to the final detection result
The staff of safety vest is worn, if it exists the staff of non-safe wearing cap and/or non-safe wearing vest, then it is assumed that should
There are security risks for construction site, and the security risk is reported monitoring center.
In embodiments of the present invention, personal safety equipment's detection model unit 81, is specifically used for:
Collect the multiple image of the construction site；
Target in each described image is labeled, markup information is generated；
The image that mark is completed is trained, personal safety equipment's detection model is generated.
In embodiments of the present invention, personal safety equipment's detection model is feedforward convolutional neural networks, the feedforward
Convolutional neural networks include: base net network and a series of convolution kernels, and the base net network is used for each frame figure in the video flowing
As carrying out feature extraction, the feature that convolution kernel is used to be extracted according to the base net network carries out target detection, to generate the target
Classification and the target position.
In embodiments of the present invention, personal safety equipment's detection model includes training loss function, the training damage
Lose the Sigmoid Classification Loss function and weight smooth L that function is weighting1The sum of loss function is positioned, the weighting
Sigmoid Classification Loss function classifies to the target in described image for optimizing personal safety equipment's model, institute
It states and weights smooth L1Positioning loss function is for optimizing the position that personal safety equipment's model determines the target.
In embodiments of the present invention, the pretreatment unit 83, is specifically used for:
Described image is zoomed in and out into preset resolution ratio.
In embodiments of the present invention, the final detection result unit 85, is specifically used for:
The testing result is screened by non-maxima suppression algorithm, obtains final detection result.
Since Installation practice and embodiment of the method under the premise of content does not conflict mutually, are filled based on same design
The content for setting embodiment can be with quoting method embodiment, and this will not be repeated here.
Referring to Fig. 9, Fig. 9 is that the embodiment of the present invention provides a kind of structural schematic diagram of intelligent terminal.Wherein, the intelligence
Energy terminal device can be personal computer, tablet computer, mobile terminal, smart phone, smartwatch etc. and can be carried out personal peace
The electronic equipment of full equipment detection.
As shown in figure 9, the intelligent terminal 90 includes one or more processors 91 and memory 92.Wherein, Fig. 9
In by taking a processor 91 as an example.
Processor 91 can be connected with memory 92 by bus or other modes, to be connected as by bus in Fig. 9
Example.
Memory 92 is used as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software journey
Sequence, non-volatile computer executable program and module, such as the detection of one of embodiment of the present invention personal safety equipment
The corresponding unit of method (for example, each unit described in Fig. 8).Processor 91 is stored in non-easy in memory 92 by operation
The property lost software program, instruction and module, thereby executing the various function application and number of the detection method of personal safety equipment
According to processing, the i.e. modules of the detection method and above-mentioned apparatus embodiment of realization above method embodiment personal safety equipment
With the function of unit.
Memory 92 may include high-speed random access memory, can also include nonvolatile memory, for example, at least
One disk memory, flush memory device or other non-volatile solid state memory parts.In some embodiments, memory 92
Optional includes the memory remotely located relative to processor 91, these remote memories can pass through network connection to processor
91.The example of above-mentioned network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
The module is stored in the memory 92, when being executed by one or more of processors 91, is executed
The detection method of personal safety equipment in above-mentioned any means embodiment, for example, executing shown in Fig. 3,4,7 described above
Each step；It can also realize the function of modules described in Fig. 8 or unit.
The intelligent terminal of the embodiment of the present invention exists in a variety of forms, including but not limited to:
(1) mobile communication equipment: the characteristics of this kind of equipment is that have mobile communication function, and to provide speech, data
Communication is main target.This class of electronic devices includes: smart phone (such as iPhone), multimedia handset, functional mobile phone, with
And low-end mobile phone etc..
(2) mobile personal computer equipment: this kind of equipment belongs to the scope of personal computer, there is calculating and processing function,
Generally also has mobile Internet access characteristic.This class of electronic devices includes: PDA, MID and UMPC equipment etc., such as iPad.
(3) portable entertainment device: this kind of equipment can show and play video content, generally also have mobile Internet access spy
Property.Such equipment includes: video player, handheld device and intelligent toy and portable car-mounted navigation equipment.
(4) other electronic equipments with video playback capability and function of surfing the Net.
The embodiment of the invention also provides a kind of nonvolatile computer storage media, the computer storage medium storage
There are computer executable instructions, which is executed by one or more processors, such as at one in Fig. 9
Device 91 is managed, may make said one or multiple processors that the inspection of the personal safety equipment in above-mentioned any means embodiment can be performed
Survey method, for example, the detection method of the personal safety equipment in above-mentioned any means embodiment is executed, for example, executing above retouch
Each step shown in Fig. 3,4,7 stated；It can also realize the function of each unit described in Fig. 8.
Device or apparatus embodiments described above is only schematical, wherein it is described as illustrated by the separation member
Unit module may or may not be physically separated, and the component shown as modular unit can be or can also
Not to be physical unit, it can it is in one place, or may be distributed on multiple network module units.It can basis
It is actual to need that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It is realized by the mode of software plus general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, above-mentioned technology
Scheme substantially in other words can be embodied in the form of software products the part that the relevant technologies contribute, the computer
Software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including some instructions are with directly
To computer equipment (can be personal computer, server or the network equipment etc.) execute each embodiment or
Method described in certain parts of embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；At this
It under the thinking of invention, can also be combined between the technical characteristic in above embodiments or different embodiment, step can be with
It is realized with random order, and there are many other variations of different aspect present invention as described above, for simplicity, they do not have
Have and is provided in details；Although the present invention is described in detail referring to the foregoing embodiments, the ordinary skill people of this field
Member is it is understood that it is still possible to modify the technical solutions described in the foregoing embodiments, or to part of skill
Art feature is equivalently replaced；And these are modified or replaceed, each reality of the application that it does not separate the essence of the corresponding technical solution
Apply the range of a technical solution.
Claims (10)
1. a kind of detection method of personal safety equipment, which is characterized in that the described method includes:
Training personal safety equipment's detection model in advance；
Obtain the video flowing of construction site；
Each frame image in the video flowing is pre-processed；
Pretreated each frame image is inputted into personal safety equipment's detection model, generates Preliminary detection result；
The Preliminary detection result is post-processed, final detection result is generated.
2. the method according to claim 1, wherein the method also includes:
According to the final detection result, the work people of non-safe wearing cap and/or non-safe wearing vest is judged whether there is
Member, the if it exists staff of non-safe wearing cap and/or non-safe wearing vest, then it is assumed that there are safety winds for the construction site
Danger, reports monitoring center for the security risk.
3. the method according to claim 1, wherein preparatory trained personal safety equipment's detection model, tool
Body includes:
Collect the multiple image of the construction site；
Target in each described image is labeled, markup information is generated；
The image that mark is completed is trained, personal safety equipment's detection model is generated.
4. the method according to claim 1, wherein personal safety equipment's detection model is feedforward convolution mind
Through network, the feedforward convolutional neural networks include: base net network and a series of convolution kernels, and the base net network is used for the video
Each frame image in stream carries out feature extraction, and the feature that convolution kernel is used to be extracted according to the base net network carries out target detection,
To generate the classification of the target and the position of the target.
5. according to the method described in claim 4, it is characterized in that, personal safety equipment's detection model includes training loss
Function, the trained loss function are the Sigmoid Classification Loss function of the weighting L smooth with weighting1Position loss function it
Sigmoid Classification Loss function with, the weighting is for optimizing personal safety equipment's model to the mesh in described image
Mark is classified, the smooth L of the weighting1Positioning loss function determines the mesh for optimizing personal safety equipment's model
Target position.
6. the method according to claim 1, wherein each frame image in the video flowing carries out in advance
Processing, comprising:
Described image is zoomed in and out into preset resolution ratio.
7. being given birth to the method according to claim 1, wherein described post-process the Preliminary detection result
At final detection result, specifically include:
The Preliminary detection result is screened by non-maxima suppression algorithm, generates final detection result.
8. a kind of detection device of personal safety equipment, it is applied to intelligent terminal, which is characterized in that described device includes:
Personal safety equipment's detection model unit, for training personal safety equipment's detection model in advance；
Video streaming units, for obtaining the video flowing of construction site；
Pretreatment unit, for being pre-processed to each frame image in the video flowing；
Preliminary detection result unit, for pretreated each frame image to be inputted personal safety equipment's detection model,
Generate Preliminary detection result；
Final detection result unit generates final detection result for post-processing to the Preliminary detection result；
Alarm Unit, for judging whether there is non-safe wearing cap and/or non-safe wearing according to the final detection result
The staff of vest, the if it exists staff of non-safe wearing cap and/or non-safe wearing vest, then it is assumed that the construction is existing
There are security risks for field, and the security risk is reported monitoring center.
9. a kind of intelligent terminal characterized by comprising
At least one processor；And
The memory being connect at least one described processor communication；Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
It manages device to execute, so that at least one described processor is able to carry out the described in any item methods of claim 1-7.
10. a kind of safety monitoring system, which is characterized in that the system comprises:
Intelligent terminal as claimed in claim 9, the intelligent terminal integrates photographic device, locality connection peripheral hardware is taken the photograph
As head or pass through the IP Camera of internet communication；
The monitoring center for passing through internet communication with the intelligent terminal；And
Receive the user terminal of the monitoring center push security risk.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910347733.5A CN110135476A (en) | 2019-04-28 | 2019-04-28 | A kind of detection method of personal safety equipment, device, equipment and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910347733.5A CN110135476A (en) | 2019-04-28 | 2019-04-28 | A kind of detection method of personal safety equipment, device, equipment and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110135476A true CN110135476A (en) | 2019-08-16 |
Family
ID=67575378
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910347733.5A Pending CN110135476A (en) | 2019-04-28 | 2019-04-28 | A kind of detection method of personal safety equipment, device, equipment and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110135476A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191581A (en) * | 2019-12-27 | 2020-05-22 | 深圳供电局有限公司 | Safety helmet detection method and device based on electric power construction and computer equipment |
| CN111753705A (en) * | 2020-06-19 | 2020-10-09 | 神思电子技术股份有限公司 | Detection method for intelligent construction site safety operation based on video analysis |
| CN111931661A (en) * | 2020-08-12 | 2020-11-13 | 桂林电子科技大学 | Real-time mask wearing detection method based on convolutional neural network |
| CN112861751A (en) * | 2021-02-22 | 2021-05-28 | 中国中元国际工程有限公司 | Airport luggage room personnel management method and device |
| 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 |
| CN113537086A (en) * | 2021-07-20 | 2021-10-22 | 科大讯飞股份有限公司 | Image recognition method and device, electronic equipment and storage medium |
| CN113627406A (en) * | 2021-10-12 | 2021-11-09 | 南方电网数字电网研究院有限公司 | Abnormal behavior detection method and device, computer equipment and storage medium |
| CN116503815A (en) * | 2023-06-21 | 2023-07-28 | 宝德计算机系统股份有限公司 | Big data-based computer vision processing system |
| CN116935473A (en) * | 2023-07-28 | 2023-10-24 | 山东智和创信息技术有限公司 | Real-time detection method and system for wearing safety helmet based on improved YOLO v7 under complex background |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108229268A (en) * | 2016-12-31 | 2018-06-29 | 商汤集团有限公司 | Expression Recognition and convolutional neural networks model training method, device and electronic equipment |
| CN108769821A (en) * | 2018-05-25 | 2018-11-06 | 广州虎牙信息科技有限公司 | Scene of game describes method, apparatus, equipment and storage medium |
| CN109255298A (en) * | 2018-08-07 | 2019-01-22 | 南京工业大学 | Safety cap detection method and system in a kind of dynamic background |
| CN109299688A (en) * | 2018-09-19 | 2019-02-01 | 厦门大学 | Ship Detection based on deformable fast convolution neural network |
| CN109447168A (en) * | 2018-11-05 | 2019-03-08 | 江苏德劭信息科技有限公司 | A kind of safety cap wearing detection method detected based on depth characteristic and video object |
-
2019
- 2019-04-28 CN CN201910347733.5A patent/CN110135476A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229268A (en) * | 2016-12-31 | 2018-06-29 | 商汤集团有限公司 | Expression Recognition and convolutional neural networks model training method, device and electronic equipment |
| CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108769821A (en) * | 2018-05-25 | 2018-11-06 | 广州虎牙信息科技有限公司 | Scene of game describes method, apparatus, equipment and storage medium |
| CN109255298A (en) * | 2018-08-07 | 2019-01-22 | 南京工业大学 | Safety cap detection method and system in a kind of dynamic background |
| CN109299688A (en) * | 2018-09-19 | 2019-02-01 | 厦门大学 | Ship Detection based on deformable fast convolution neural network |
| CN109447168A (en) * | 2018-11-05 | 2019-03-08 | 江苏德劭信息科技有限公司 | A kind of safety cap wearing detection method detected based on depth characteristic and video object |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191581A (en) * | 2019-12-27 | 2020-05-22 | 深圳供电局有限公司 | Safety helmet detection method and device based on electric power construction and computer 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 |
| CN111753705B (en) * | 2020-06-19 | 2023-11-03 | 神思电子技术股份有限公司 | Intelligent construction site safety operation detection method based on video analysis |
| CN111753705A (en) * | 2020-06-19 | 2020-10-09 | 神思电子技术股份有限公司 | Detection method for intelligent construction site safety operation based on video analysis |
| CN111931661A (en) * | 2020-08-12 | 2020-11-13 | 桂林电子科技大学 | Real-time mask wearing detection method based on convolutional neural network |
| CN112861751A (en) * | 2021-02-22 | 2021-05-28 | 中国中元国际工程有限公司 | Airport luggage room personnel management method and device |
| CN112861751B (en) * | 2021-02-22 | 2024-01-12 | 中国中元国际工程有限公司 | Airport luggage room personnel management method and device |
| CN113537086A (en) * | 2021-07-20 | 2021-10-22 | 科大讯飞股份有限公司 | Image recognition method and device, electronic equipment and storage medium |
| CN113627406A (en) * | 2021-10-12 | 2021-11-09 | 南方电网数字电网研究院有限公司 | Abnormal behavior detection method and device, computer equipment and storage medium |
| CN113627406B (en) * | 2021-10-12 | 2022-03-08 | 南方电网数字电网研究院有限公司 | Abnormal behavior detection method and device, computer equipment and storage medium |
| CN116503815A (en) * | 2023-06-21 | 2023-07-28 | 宝德计算机系统股份有限公司 | Big data-based computer vision processing system |
| CN116503815B (en) * | 2023-06-21 | 2024-01-30 | 宝德计算机系统股份有限公司 | Big data-based computer vision processing system |
| CN116935473A (en) * | 2023-07-28 | 2023-10-24 | 山东智和创信息技术有限公司 | Real-time detection method and system for wearing safety helmet based on improved YOLO v7 under complex background |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110135476A (en) | A kind of detection method of personal safety equipment, device, equipment and system | |
| CN110147743A (en) | Real-time online pedestrian analysis and number system and method under a kind of complex scene | |
| CN106469302A (en) | A kind of face skin quality detection method based on artificial neural network | |
| CN103268495B (en) | Human body behavior modeling recognition methods based on priori knowledge cluster in computer system | |
| CN109919251A (en) | A kind of method and device of object detection method based on image, model training | |
| CN110096933A (en) | The method, apparatus and system of target detection | |
| CN111626116B (en) | Video semantic analysis method based on fusion of multi-attention mechanism and Graph | |
| CN108921039A (en) | The forest fire detection method of depth convolution model based on more size convolution kernels | |
| CN108470138A (en) | Method for target detection and device | |
| CN107423398A (en) | Exchange method, device, storage medium and computer equipment | |
| CN106462725A (en) | Systems and methods of monitoring activities at a gaming venue | |
| CN109214366A (en) | Localized target recognition methods, apparatus and system again | |
| CN110516529A (en) | It is a kind of that detection method and system are fed based on deep learning image procossing | |
| CN110084165A (en) | The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations | |
| Wang et al. | Investigation into recognition algorithm of helmet violation based on YOLOv5-CBAM-DCN | |
| CN112966589A (en) | Behavior identification method in dangerous area | |
| CN104063686A (en) | System and method for performing interactive diagnosis on crop leaf segment disease images | |
| CN110263654A (en) | A kind of flame detecting method, device and embedded device | |
| CN104537273B (en) | A kind of drowned pattern intelligent inference system and method | |
| CN111914676A (en) | Human body tumbling detection method and device, electronic equipment and storage medium | |
| CN109903339A (en) | A kind of video group personage's position finding and detection method based on multidimensional fusion feature | |
| CN111368768A (en) | Human body key point-based employee gesture guidance detection method | |
| CN110399822A (en) | Action identification method of raising one's hand, device and storage medium based on deep learning | |
| CN110532838A (en) | Object test equipment and method and storage medium | |
| Yandouzi et al. | Investigation of combining deep learning object recognition with drones for forest fire detection and monitoring |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190816 |
|
| RJ01 | Rejection of invention patent application after publication |