CN111325132A - Intelligent monitoring system - Google Patents
Intelligent monitoring system Download PDFInfo
- Publication number
- CN111325132A CN111325132A CN202010095650.4A CN202010095650A CN111325132A CN 111325132 A CN111325132 A CN 111325132A CN 202010095650 A CN202010095650 A CN 202010095650A CN 111325132 A CN111325132 A CN 111325132A
- Authority
- CN
- China
- Prior art keywords
- human body
- image
- judged
- safe
- face
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 66
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 58
- 238000012806 monitoring device Methods 0.000 claims abstract description 7
- 230000001815 facial effect Effects 0.000 claims description 25
- 238000000034 method Methods 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 238000001514 detection method Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 210000000697 sensory organ Anatomy 0.000 claims description 7
- 238000013135 deep learning Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000003287 optical effect Effects 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 210000001508 eye Anatomy 0.000 description 3
- 210000004709 eyebrow Anatomy 0.000 description 3
- 210000001331 nose Anatomy 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 210000000088 lip Anatomy 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- 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
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an intelligent monitoring system, which comprises the following equipment: target monitoring equipment for monitoring a target existing in the warning area; the image positioning device is used for positioning the target object and acquiring a group of images to be judged within a certain time after the target object monitoring device identifies the target object; the human body identification device is used for acquiring the group of images to be judged from the image positioning device and identifying whether human bodies exist in the group of images to be judged; the target identification device is used for identifying one image to be judged with the most human body characteristics in the group of images to be judged as a human body image to be detected when the human body identification device judges that human bodies exist in the group of images to be judged; then, acquiring a safe human body image from a background database; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the human body image to be detected is a safe human body result or not according to the safe human body characteristics.
Description
Technical Field
The invention relates to the technical field of monitoring equipment, in particular to an intelligent monitoring system.
Background
Currently, monitoring of objects in an area, especially image recognition and monitoring, is an important task and need. However, if the monitoring is completely dependent on manual monitoring, firstly, the efficiency is low, secondly, the cost is increased, thirdly, the monitoring error rate or the leakage rate is high, and the defects are always puzzled on enterprises with related monitoring requirements.
In recent years, technologies and applications based on computer image recognition are increasingly emerging, and the application of computer vision technologies, such as image recognition, to intelligent monitoring is an important technical development idea, and how to better optimize and utilize image recognition to be applied to the monitoring field and how to optimize the monitoring process to improve accuracy rate are one of the technical problems to be solved in the related technical field.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention solves the technical problem of an intelligent monitoring system which can optimize the monitoring process and improve the accuracy.
In order to solve the technical problems, the technical scheme adopted by the invention specifically comprises the following contents:
an intelligent monitoring system, comprising the following devices:
target monitoring equipment for monitoring a target existing in the warning area;
the image positioning device is used for positioning the target object and acquiring a group of images to be judged within a certain time after the target object monitoring device identifies the target object;
the human body identification device is used for acquiring the group of images to be judged from the image positioning device and identifying whether human bodies exist in the group of images to be judged;
the target identification device is used for identifying one image to be judged with the most human body characteristics in the group of images to be judged as a human body image to be detected when the human body identification device judges that human bodies exist in the group of images to be judged; then, acquiring a safe human body image from a background database; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the safe human body image is a safe human body result or not according to the safe human body characteristics.
In order to optimize the monitoring process and improve the accuracy, on one hand, the inventor innovatively proposes to identify and acquire images aiming at a target object existing in the warning area in the technical scheme, and the acquired images are a group of images to be judged acquired within a certain time instead of only one image, so that unnecessary identification of people mistakenly entering the warning area within a very short time can be effectively avoided. That is, the identification only needs to perform subsequent identification work on the target objects continuously appearing in the warning area within a period of time, unnecessary work of equipment identification is reduced, and therefore the working frequency of the equipment is reduced, and the equipment is turned to only identify effective and really needed monitoring identification.
It should be noted that the certain time period may be 5s, 10s or 1min, which may be set according to the time when the actual warning area enters mistakenly.
The above-mentioned continuity may be continuity when the frequency of occurrence reaches a certain amount within the above-mentioned certain time, and does not need to occur at all times within the certain time. For example, if a certain time is set to 5s, the requirement of monitoring and identification is met when 3s appears in the 5s, that is, the target object can be located and a group of images to be determined can be acquired.
On the other hand, the inventor also innovatively learns the safe human body characteristics by utilizing the safe human body images in the background database through the deep convolutional network and judges whether the human body image to be detected is a safe human body according to the safe human body characteristics so as to determine whether the target object is safe. The deep convolutional neural network is trained, so that various characteristic information of pedestrians can be integrated, the deep convolutional network is utilized to learn human body characteristics, various factors influencing the recognition accuracy can be reduced, such as the visual angle, the resolution ratio and the like of an image, and the technical purpose of more accurate recognition matching is achieved.
It should be noted that the secure human body refers to human body information that has been authenticated, and is monitored without further processing.
Preferably, the human body recognition device further comprises the following steps: if the human bodies exist in the group of images to be judged, judging whether the same number of human bodies exist in each image to be judged in the group;
if so, the target identification equipment identifies one image to be judged with the most human body features in the group of images to be judged as a human body image to be detected;
if not, firstly selecting a plurality of images to be judged with the largest number and consistency of human bodies, and then identifying one image to be judged with the largest human body characteristics in the images to be judged as the image of the human body to be judged.
It should be noted that, because a group of images to be determined have different information such as human body positions, and therefore have different human body characteristics, in order to perform better identification, in the group of images to be determined, the target identification device selects one image to be determined having the most human body characteristics as the image of the human body to be detected for subsequent monitoring identification, which can improve the accuracy of identification.
Preferably, the target recognition device further includes a face recognition step, specifically including:
and identifying first face information of the human body image to be detected by using a face identification module, matching the first face information with second face information of the safe human body image, and judging whether the safe human body image is a safe human body.
On one hand, the accuracy of identification is improved, and the accuracy of judgment by using the deep convolutional neural network is assisted; on the other hand, in order to prevent the situation that the whole intelligent monitoring system cannot recognize due to the fact that the deep convolutional neural network cannot recognize under certain conditions, the inventor adds a scheme of obtaining the face information of the to-be-recognized human body image and matching the face information with the face information of the safe human body image in a preferred implementation mode of the technical scheme, and can effectively recognize the to-be-judged image from another angle, so that the operation effectiveness of the whole intelligent monitoring system is improved.
More preferably, the target recognition device further includes, before recognizing the first face information of the human body image to be measured, the steps of:
dividing the image to be judged into a first face area and a first image area to be judged; dividing the safe human body image into a second face area and a second background area; the color of the second background area is then overlaid over the first background area.
It should be noted that, in order to improve the accuracy and effectiveness of the technical scheme of face recognition, the inventor adds an optimized technical scheme for denoising the background region in a further technical scheme, that is, both the image to be judged and the safe human body image are divided into the face region and the background region, and the image to be judged of the image to be judged is replaced with the background region of the safe human body image, thereby improving the accuracy of face recognition. In addition, the techniques used to divide the face region and the background region are well known to those skilled in the art and will not be described herein.
More preferably, the method of the face recognition step specifically comprises:
training a face recognition module;
obtaining the human body image to be detected, detecting the human body image to be detected by using a face recognition module obtained by training to obtain a face area rl in the human body image to be detected, and positioning the five sense organs in the face area rl to obtain first five sense organs data of the human body image to be detected;
determining a safe human face image, and carrying out facial feature detection on the safe human face image to obtain second facial feature data of the safe human face image;
calculating the similarity of the whole face and the local facial features of the first facial feature and the second facial feature based on the data of the first facial feature and the data of the second facial feature;
and calculating the probability fusion similarity of the human body image to be detected and the safe human face image to obtain a judgment result.
It should be noted that the face recognition module obtained through training can obtain relevant data information in advance, so that subsequent recognition and matching are facilitated, and the operation efficiency is improved, thereby improving the work and operation efficiency of the whole intelligent monitoring system.
Further, if the acquired image of the human body to be detected is acquired, the human face recognition module detects the image of the human body to be detected by using an ASM algorithm; if the group of images to be judged is obtained, unsupervised learning is adopted, namely a CNN model is used for detecting the positions of the key points of the human face, and then optical flow (flow tracking) is used for tracking the positions of the key points of the human face of the next frame of image of the group of images to be judged to serve as fusion information to a CNN detector to serve as auxiliary information of the key points of the human face.
Preferably, the object recognition device further comprises an input device; the target recognition device is further used for processing the safe human body description through natural language and screening out safe human body key information after the safe human body description is obtained from the input device, and then obtaining safe human body characteristics by applying deep learning through the safe human body key information.
It should be noted that the input device may be a computer interactive system or the like. The human body description can be the physical sign of the human body description, and the human body description can be screened out from human body key information such as height and the like by natural language processing of the system after being input.
More preferably, the deep convolutional neural network is constructed by using a CNN model pre-trained by ImageNet using an AlexNet classical network, and is obtained by training all safe human body images in the background database.
It should be noted that the AlexNet network model has numerous parameters, and in a preferred embodiment, if the secure human body data set required in the system monitoring process is numerous, for example, an enterprise with thousands of people, it is more suitable to apply the AlexNet classical network, which can obtain better learning results, and through the learning of numerous pictures, interference factors can be better eliminated, effective human body characteristics can be obtained, and thus better matching results can be realized.
More preferably, before the target recognition device obtains the safe human body image from the background database, the method further comprises a preliminary recognition step: carrying out primary identification on a human body image to be detected by using a first convolution neural network, and acquiring a safe human body image from a background database if the result of the primary identification is that a suspicious target exists in the human body image to be detected; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the safe human body image is a safe human body result or not according to the safe human body characteristics.
Further, the first convolutional neural network and the deep convolutional neural network are obtained by training different small batch data sets of a background database through the same model, wherein the deep convolutional neural network is larger in size than the first convolutional neural network.
Still further, the first convolutional neural network is a TensorFlow-based small convolutional neural network.
It should be noted that the first convolutional neural network may adopt a tensrflow-based small convolutional neural network. In the preferred embodiment, the TensorFlow small convolutional neural network is adopted for preliminary identification before identification, a part of suspicious personnel can be excluded in advance, the TensorFlow adopted for preliminary identification has the greatest advantage of lower learning threshold, deep learning is not needed like a deep convolutional neural network, the process is simpler, subsequent deep convolutional neural network identification is not needed after the preliminary identification is excluded, the safety result is directly returned, the efficiency is improved, and the learning cost is reduced.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the intelligent monitoring system, the subsequent identification work is only needed to be carried out on the target objects which continuously appear in the warning area within a period of time, so that the unnecessary work of equipment identification is reduced, the working frequency of the equipment is reduced, and the equipment is turned to only identify effective and really needed monitoring identification;
2. the intelligent monitoring system learns the safety human body characteristics by utilizing the safety human body images in the background database through the deep convolutional network, judges whether the human body image to be detected is a safety human body according to the safety human body characteristics, and determines whether the target object is safe. The deep convolutional neural network is trained, so that various characteristic information of pedestrians can be integrated, the deep convolutional network is utilized to learn human body characteristics, and various factors influencing the recognition accuracy, such as the visual angle, the resolution ratio and the like of an image, can be reduced, so that the technical purpose of more accurate recognition matching is realized;
3. the intelligent monitoring system of the invention adds a scheme of acquiring the face information of the human body image to be detected and matching the face information with the face information of the safe human body image, and can effectively identify the image to be judged from another angle, thereby improving the operation effectiveness of the whole intelligent monitoring system;
4. the intelligent monitoring system of the invention adopts an optimized technical scheme for denoising a background region, namely, dividing an image to be judged and a safe human body image into a face region and a background region, and replacing the image to be judged of the image to be judged with the background region of the safe human body image, thereby improving the accuracy of face recognition;
5. the intelligent monitoring system adopts the TensorFlow small-sized convolutional neural network for preliminary identification before identification, can exclude a part of suspicious personnel in advance, has the biggest advantage of lower learning threshold without deep learning like the deep convolutional neural network, has simpler process, does not need subsequent deep convolutional neural network identification after the preliminary identification is eliminated, directly returns a safety result, improves the efficiency and reduces the learning cost.
The foregoing description is only an overview of the technical solutions of the present invention, and in order to make the technical means of the present invention more clearly understood, the present invention may be implemented in accordance with the content of the description, and in order to make the above and other objects, features, and advantages of the present invention more clearly understood, the following preferred embodiments are described in detail with reference to the accompanying drawings.
Drawings
FIG. 1 is a flow chart of the operation of a preferred embodiment of the intelligent monitoring system of the present invention;
FIG. 2 is a diagram of the connection relationship of the devices in another preferred embodiment of the intelligent monitoring system of the present invention;
fig. 3 is a device connection diagram of another preferred embodiment of the intelligent monitoring system of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the embodiments, structures, features and effects according to the present invention with reference to the accompanying drawings and preferred embodiments is as follows:
fig. 1 is a flow chart of a preferred embodiment of the intelligent monitoring system of the present invention, which describes a flow of the intelligent monitoring system, wherein the intelligent monitoring device comprises the following devices:
target monitoring equipment for monitoring a target existing in the warning area; in some specific embodiments, the object monitoring device may be a camera or other specific device structure.
The image positioning device is used for positioning the target object and acquiring a group of images to be judged within a certain time after the target object monitoring device identifies the target object;
the human body identification device is used for acquiring the group of images to be judged from the image positioning device and identifying whether human bodies exist in the group of images to be judged;
the target identification device is used for identifying one image to be judged with the most human body characteristics in the group of images to be judged as a human body image to be detected when the human body identification device judges that human bodies exist in the group of images to be judged;
the method also comprises a preliminary identification step before acquiring the safe human body image from the background database: carrying out primary identification on a human body image to be detected by using a first convolution neural network, and acquiring a safe human body image from a background database if the result of the primary identification is that a suspicious target exists in the human body image to be detected; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the safe human body image is a safe human body result or not according to the safe human body characteristics. The first convolutional neural network is a TensorFlow-based small convolutional neural network.
The device connection diagram of the above device can be seen with reference to fig. 2 and 3, where fig. 3 is the device state when the preliminary identification step is provided, and fig. 2 is the device state when the preliminary identification step is not provided. As can be seen from the figure, the target object monitoring device may be implemented by a specific device structure such as a camera, for example, a binocular camera; the image positioning device, the human body recognition device and the target recognition device can be integrated in a terminal to be realized, wherein the information of the image positioning device can be obtained through pictures which are monitored and shot by the camera. In a specific embodiment, the terminal device may be a personal computer or the like. In addition, the terminal equipment is also provided with a result returning module for returning the identification result. The result returning module can be connected with a mobile phone, a monitor or an alarm system of a monitoring person in an extensible manner so as to realize the subsequent alarm monitoring effect. And the background server is provided with a data set in which safe human body information is stored.
In this embodiment, on one hand, the inventor innovatively proposes to identify and acquire images of a target object existing in the warning area, and the acquired images are a group of images to be determined acquired within a certain time, rather than acquiring only one image, so that unnecessary identification of a person who mistakenly enters the warning area within a very short time can be effectively avoided. That is, the identification only needs to perform subsequent identification work on the target objects continuously appearing in the warning area within a period of time, unnecessary work of equipment identification is reduced, and therefore the working frequency of the equipment is reduced, and the equipment is turned to only identify effective and really needed monitoring identification. On the other hand, the inventor also innovatively learns the safe human body characteristics by utilizing the safe human body images in the background database through the deep convolutional network and judges whether the human body image to be detected is a safe human body according to the safe human body characteristics so as to determine whether the target object is safe. The deep convolutional neural network is trained, so that various characteristic information of pedestrians can be integrated, the deep convolutional network is utilized to learn human body characteristics, various factors influencing the recognition accuracy can be reduced, such as the visual angle, the resolution ratio and the like of an image, and the technical purpose of more accurate recognition matching is achieved.
From the above two recognition processes, on the one hand, the first convolutional neural network and the deep neural network can be obtained through learning in advance. Therefore, the learnt neural network model can be used for rapidly identifying the human body characteristics of the object to be detected, and the intelligent monitoring efficiency can be improved; on the other hand, the target object is confirmed to be the template needing alarm monitoring through two times of identification, and the identification accuracy of intelligent monitoring is improved.
In some more specific embodiments, the certain time period may be 5s, 10s or 1min, which may be set according to the time when the actual warning area is mistakenly entered.
In the embodiment, the TensorFlow small-sized convolutional neural network is adopted for primary identification before identification, a part of suspicious personnel can be excluded in advance, the TensorFlow adopted for primary identification has the greatest advantage that the learning threshold is lower, deep learning is not needed like a deep convolutional neural network, the process is simpler, subsequent deep convolutional neural network identification is not needed after the primary identification is excluded, the safety result is directly returned, the efficiency is improved, and the learning cost is reduced. In other embodiments, a preliminary identification step may be provided depending on the actual situation.
With reference to the foregoing embodiment, in another preferred embodiment, the human body identification device further includes the following steps: if the human bodies exist in the group of images to be judged, judging whether the same number of human bodies exist in each image to be judged in the group; if so, the target identification equipment identifies one image to be judged with the most human body features in the group of images to be judged as a human body image to be detected; if not, firstly selecting a plurality of images to be judged with the largest number and consistency of human bodies, and then identifying one image to be judged with the largest human body characteristics in the images to be judged as the image of the human body to be judged.
Because information such as human body positions of a group of images to be judged is different, human body characteristics contained in the images to be judged are also different, in order to better identify, in the group of images to be judged, the target identification equipment selects one image to be judged with the most human body characteristics as the image of the human body to be detected to carry out subsequent monitoring and identification, and the identification accuracy can be improved.
With reference to the foregoing embodiment, in another preferred embodiment, the object recognition device further includes:
and identifying first face information of the human body image to be detected, matching the first face information with second face information of the safe human body image, and judging whether the safe human body image is a safe human body.
In this embodiment, on one hand, to improve the accuracy of identification and assist the accuracy of the judgment by using the deep convolutional neural network; on the other hand, in order to prevent the situation that the whole intelligent monitoring system cannot recognize due to the fact that the deep convolutional neural network cannot recognize under certain conditions, the inventor adds a scheme of obtaining the face information of the to-be-recognized human body image and matching the face information with the face information of the safe human body image in a preferred implementation mode of the technical scheme, and can effectively recognize the to-be-judged image from another angle, so that the operation effectiveness of the whole intelligent monitoring system is improved.
With reference to the foregoing embodiment, in another preferred embodiment, before identifying the first face information of the image of the human body to be detected, the target identification device further includes the following steps:
dividing the image to be judged into a first face area and a first image area to be judged; dividing the safe human body image into a second face area and a second background area; the color of the second background area is then overlaid over the first background area.
In order to improve the accuracy and effectiveness of the technical scheme of face recognition, the inventor adds an optimization technical scheme for denoising a background region in a further technical scheme, namely, dividing an image to be judged and a safe human body image into the face region and the background region, and replacing the image to be judged of the image to be judged with the background region of the safe human body image, thereby improving the accuracy of face recognition. In addition, the techniques used to divide the face region and the background region are well known to those skilled in the art and will not be described herein.
In some specific embodiments, the method for face recognition may employ the following steps:
s1: training a face recognition module;
s2: obtaining the human body image to be detected, detecting the human body image to be detected by using a face recognition module obtained by training to obtain a face area rl in the human body image to be detected, and positioning the five sense organs in the face area rl to obtain first five sense organs data of the human body image to be detected;
s3: determining a safe human face image, and carrying out facial feature detection on the safe human face image to obtain second facial feature data of the safe human face image;
s4: calculating the similarity of the whole face and the local facial features of the two facial features based on the first facial feature data obtained in the step S2 and the second facial feature data obtained in the step S3;
and calculating the probability fusion similarity of the human body image to be detected and the safe human face image to obtain a judgment result.
In some more specific embodiments, the data set utilized by the training face recognition module may be the data set of the background database, or may be other data sets. The number of pictures in the data set can be thousands of units, and the same person can adopt face pictures with different angles and illumination forms, so that the user can learn more accurately. During learning training, a test set may be established for testing the generalization level of the trained model. In the training process, one-stage algorithms such as Yolo and SSD can be adopted, the algorithm has the advantages that the algorithm for directly predicting the types and the positions of different targets by using only one CNN network is high in speed, the training and detecting speed and efficiency can be improved, and compared with other training algorithms, the one-stage algorithms are particularly suitable for face recognition.
In combination with the above embodiments, in some other more specific embodiments, the use of ASM detection methods; on the basis, unsupervised learning can be combined, namely, a CNN model is adopted to detect the positions of the face key points, and then optical flow (flow tracking) is used for tracking the positions of the face key points of the next frame of picture of the group of images to be judged to be used as fusion information to be sent to a CNN detector to be used as auxiliary information of the face key points so as to detect the first facial feature data of the face region rl. The ASM detection method can be used for the face detection of more traditional and scattered pictures, and the CNN model can be used for detecting the pictures of continuous frames of the group of images to be judged to perform the face detection. The specific training method of the ASM or CNN model is known by those skilled in the art, and is not considered as the point of the invention of the present patent, but is applied to the intelligent monitoring system of the present technical solution.
When the ASM detection method is adopted, the key points of the human face can be respectively arranged on the outline of the whole face and key points of facial features of the human face, such as the eyebrows, the eyes, the nose, the lips and the like, so that a key point distribution image of the facial features of the whole face is obtained; and according to the obtained key point distribution image of the five sense organs of the whole face, local images such as eyebrows, eyes, noses, lips and the like in the face area rl are cut out.
The following embodiment may be specifically adopted when calculating the full-face and local facial feature similarity of both the first and second facial feature data; first, the eyebrows, the eyes, the nose, the lips, etc. on the upper face can be taken as four key regions, and in addition, the whole face can be taken as five images in total, and the five images are respectively trained to carry out feature extraction and classification on the corresponding CNNs. Then, the probability of each category in the four key areas is calculated according to the convolutional neural network, and the probability that two faces of the four categories are judged to be the same person is output to the whole face. The technical scheme has the advantages that the local features are more detailed and stable than the overall features, the local features are not influenced by the overall expression, and the recognition accuracy is improved; the judgment is more comprehensive and accurate by combining the judgment of local and overall characteristics; in addition, when the fusion probability is calculated, the output results of all CNNs can be fused, the face features are fully utilized, the face similarity calculation of multiple CNNs based on probability fusion can be realized, and the calculation accuracy can be effectively improved.
With reference to the foregoing embodiment, in another preferred embodiment, the object recognition device further includes an input device; the target recognition device is further used for processing the safe human body description through natural language and screening out safe human body key information after the safe human body description is obtained from the input device, and then obtaining safe human body characteristics by applying deep learning through the safe human body key information.
In some more specific embodiments, the input device may be a computer interactive system or the like. The human body description can be the physical sign of the human body description, and the human body description can be screened out from human body key information such as height and the like by natural language processing of the system after being input.
In some more specific embodiments, the deep convolutional neural network is constructed using a CNN model pre-trained by ImageNet using an AlexNet classical network and is trained using all safe human body images in the background database. The AlexNet network model has numerous parameters, and in a preferred embodiment, if the safety human body data set required in the system monitoring process is numerous, for example, an enterprise with thousands of people, the AlexNet classical network is more suitable, so that a better learning result can be obtained, interference factors can be better eliminated through the learning of numerous pictures, effective human body characteristics are obtained, and a better matching result is realized.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above embodiments are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereby, and any insubstantial changes and substitutions made by those skilled in the art based on the present invention are within the protection scope of the present invention.
Claims (10)
1. An intelligent monitoring system, characterized in that, it includes following equipment:
target monitoring equipment for monitoring a target existing in the warning area;
the image positioning device is used for positioning the target object and acquiring a group of images to be judged within a certain time after the target object monitoring device identifies the target object;
the human body identification device is used for acquiring the group of images to be judged from the image positioning device and identifying whether human bodies exist in the group of images to be judged;
the target identification device is used for identifying one image to be judged with the most human body characteristics in the group of images to be judged as a human body image to be detected when the human body identification device judges that human bodies exist in the group of images to be judged; then, acquiring a safe human body image from a background database; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the human body image to be detected is a safe human body result or not according to the safe human body characteristics.
2. The intelligent monitoring system according to claim 1, wherein the human body recognition device further comprises the following steps during recognition: if the human bodies exist in the group of images to be judged, judging whether the same number of human bodies exist in each image to be judged in the group;
if so, the target identification equipment identifies one image to be judged with the most human body features in the group of images to be judged as a human body image to be detected;
if not, firstly selecting a plurality of images to be judged with the largest number and consistency of human bodies, and then identifying one image to be judged with the largest human body characteristics in the images to be judged as the image of the human body to be judged.
3. The intelligent monitoring system according to claim 1, wherein the target recognition device further comprises a face recognition step, specifically comprising:
and identifying first face information of the human body image to be detected by using a face identification module, matching the first face information with second face information of the safe human body image, and judging whether the safe human body image is a safe human body.
4. The intelligent monitoring system according to claim 3, wherein the object recognition device further comprises the following steps before recognizing the first face information of the human body image to be tested:
dividing the image to be judged into a first face area and a first background area; dividing the safe human body image into a second face area and a second background area; the color of the second background area is then overlaid over the first background area.
5. The intelligent monitoring system according to claim 3, wherein the face recognition step is specifically performed by:
training a face recognition module;
acquiring the human body image to be detected or the group of images to be judged, detecting the human body image to be detected or the group of images to be judged by using a face recognition module obtained by training to obtain a face area rl in the human body image to be detected, and positioning the five sense organs in the face area rl to obtain first five sense organ data of the human body image to be detected;
determining a safe human face image, and carrying out facial feature detection on the safe human face image to obtain second facial feature data of the safe human face image;
calculating the similarity of the whole face and the local facial features of the first facial feature and the second facial feature based on the data of the first facial feature and the data of the second facial feature;
and calculating the probability fusion similarity of the human body image to be detected and the safe human face image to obtain a judgment result.
6. The intelligent monitoring system according to claim 5, wherein if the acquired image of the human body to be detected is, the face recognition module detects the image by using an ASM algorithm; if the group of images to be judged is obtained, unsupervised learning is adopted, namely a CNN model is used for detecting the positions of the key points of the human face, and then optical flow (flow tracking) is used for tracking the positions of the key points of the human face of the next frame of image of the group of images to be judged to serve as fusion information to a CNN detector to serve as auxiliary information of the key points of the human face.
7. The intelligent monitoring system of claim 1, wherein the object-recognition device further comprises an input device; the target recognition device is further used for processing the safe human body description through natural language and screening out safe human body key information after the safe human body description is obtained from the input device, and then obtaining safe human body characteristics by applying deep learning through the safe human body key information.
8. The intelligent monitoring system according to claim 7, wherein the deep convolutional neural network is constructed using a CNN model pre-trained by AlexNet classical network in ImageNet and is trained using all safe human body images in the background database.
9. The intelligent monitoring system according to claim 7, wherein before the target recognition device obtains the safe human body image from the background database, the method further comprises a preliminary recognition step of:
carrying out primary identification on a human body image to be detected by using a first convolution neural network, and acquiring a safe human body image from a background database if the result of the primary identification is that a suspicious target exists in the human body image to be detected; and learning the safe human body characteristics from the safe human body image by utilizing a deep convolutional neural network, and judging whether the safe human body image is a safe human body result or not according to the safe human body characteristics.
10. The intelligent monitoring system according to claim 9, wherein the first convolutional neural network and the deep convolutional neural network are obtained via training the same model on different small-lot data sets of a background database, wherein the deep convolutional neural network is larger in size than the first convolutional neural network; the first convolutional neural network is a TensorFlow-based small convolutional neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010095650.4A CN111325132A (en) | 2020-02-17 | 2020-02-17 | Intelligent monitoring system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010095650.4A CN111325132A (en) | 2020-02-17 | 2020-02-17 | Intelligent monitoring system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111325132A true CN111325132A (en) | 2020-06-23 |
Family
ID=71172715
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010095650.4A Pending CN111325132A (en) | 2020-02-17 | 2020-02-17 | Intelligent monitoring system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111325132A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380987A (en) * | 2020-11-13 | 2021-02-19 | 无锡臻永科技有限公司 | Effective learning duration statistical system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108460358A (en) * | 2018-03-20 | 2018-08-28 | 武汉倍特威视系统有限公司 | Safety cap recognition methods based on video stream data |
| CN109274891A (en) * | 2018-11-07 | 2019-01-25 | 北京旷视科技有限公司 | A kind of image processing method, device and its storage medium |
| CN109522853A (en) * | 2018-11-22 | 2019-03-26 | 湖南众智君赢科技有限公司 | Face datection and searching method towards monitor video |
| CN109727350A (en) * | 2018-12-14 | 2019-05-07 | 深圳壹账通智能科技有限公司 | A kind of Door-access control method and device based on recognition of face |
-
2020
- 2020-02-17 CN CN202010095650.4A patent/CN111325132A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108460358A (en) * | 2018-03-20 | 2018-08-28 | 武汉倍特威视系统有限公司 | Safety cap recognition methods based on video stream data |
| CN109274891A (en) * | 2018-11-07 | 2019-01-25 | 北京旷视科技有限公司 | A kind of image processing method, device and its storage medium |
| CN109522853A (en) * | 2018-11-22 | 2019-03-26 | 湖南众智君赢科技有限公司 | Face datection and searching method towards monitor video |
| CN109727350A (en) * | 2018-12-14 | 2019-05-07 | 深圳壹账通智能科技有限公司 | A kind of Door-access control method and device based on recognition of face |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380987A (en) * | 2020-11-13 | 2021-02-19 | 无锡臻永科技有限公司 | Effective learning duration statistical system |
| CN112380987B (en) * | 2020-11-13 | 2021-12-17 | 圣凯诺服饰有限公司 | Effective learning duration statistical system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110210302B (en) | Multi-target tracking method, device, computer equipment and storage medium | |
| US11961227B2 (en) | Method and device for detecting and locating lesion in medical image, equipment and storage medium | |
| WO2021004112A1 (en) | Anomalous face detection method, anomaly identification method, device, apparatus, and medium | |
| CN110751022A (en) | Urban pet activity track monitoring method based on image recognition and related equipment | |
| CN111726586A (en) | Production system operation standard monitoring and reminding system | |
| CN111062239A (en) | Human body target detection method and device, computer equipment and storage medium | |
| CN109165685B (en) | Expression and action-based method and system for monitoring potential risks of prisoners | |
| CN105117692A (en) | Real-time face identification method and system based on deep learning | |
| EP3680794A1 (en) | Device and method for user authentication on basis of iris recognition | |
| CN110555875A (en) | Pupil radius detection method and device, computer equipment and storage medium | |
| CN110991397B (en) | Travel direction determining method and related equipment | |
| CN111062303A (en) | Image processing method, system and computer storage medium | |
| CN111325133B (en) | Image processing system based on artificial intelligent recognition | |
| Dzhangarov et al. | Face recognition methods | |
| CN111415339A (en) | Image defect detection method for complex texture industrial product | |
| CN116863522A (en) | Acne grading method, device, equipment and medium | |
| CN111353385B (en) | Pedestrian re-identification method and device based on mask alignment and attention mechanism | |
| CN112307984A (en) | Safety helmet detection method and device based on neural network | |
| Putra | A Novel Method for Handling Partial Occlusion on Person Re-identification using Partial Siamese Network | |
| CN107862298A (en) | It is a kind of based on the biopsy method blinked under infrared eye | |
| CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
| CN114359787A (en) | Target attribute identification method and device, computer equipment and storage medium | |
| CN111325132A (en) | Intelligent monitoring system | |
| Wang et al. | A deep learning-based method for vehicle licenseplate recognition in natural scene | |
| CN115273150A (en) | Novel identification method and system for wearing safety helmet based on human body posture estimation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200623 |