CN112651347A - Smoking behavior sample generation method and system based on double-spectrum imaging - Google Patents

Smoking behavior sample generation method and system based on double-spectrum imaging Download PDF

Info

Publication number
CN112651347A
CN112651347A CN202011593064.9A CN202011593064A CN112651347A CN 112651347 A CN112651347 A CN 112651347A CN 202011593064 A CN202011593064 A CN 202011593064A CN 112651347 A CN112651347 A CN 112651347A
Authority
CN
China
Prior art keywords
cigarette end
target point
end target
visible light
point
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.)
Granted
Application number
CN202011593064.9A
Other languages
Chinese (zh)
Other versions
CN112651347B (en
Inventor
王勇
周子誉
王春林
李倩
单宝旭
李畅昊
赵宏
孙俊杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiaxing Hengchuang Electric Power Group Co ltd Bochuang Material Branch
Jiaxing Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
Jiaxing Hengchuang Electric Power Group Co ltd Bochuang Material Branch
Jiaxing Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiaxing Hengchuang Electric Power Group Co ltd Bochuang Material Branch, Jiaxing Power Supply Co of State Grid Zhejiang Electric Power Co Ltd filed Critical Jiaxing Hengchuang Electric Power Group Co ltd Bochuang Material Branch
Priority to CN202011593064.9A priority Critical patent/CN112651347B/en
Publication of CN112651347A publication Critical patent/CN112651347A/en
Application granted granted Critical
Publication of CN112651347B publication Critical patent/CN112651347B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a method and a system for generating smoking behavior samples based on double-spectrum imaging, wherein the method comprises the following steps: performing pixel matching on a visible light camera and an infrared imaging camera to obtain a homography transformation matrix; acquiring a visible light image based on a visible light camera, and carrying out single skeleton detection on people in the visible light image to obtain a human body trunk key point of each person; acquiring an infrared image based on an infrared imaging camera, carrying out high-temperature point detection on the infrared image, determining a cigarette end target point in the infrared image, and calculating pixel coordinates of the cigarette end target point in a visible light image through a homography transformation matrix; matching figures and cigarette end target points; and marking the minimum outer envelope rectangle of the key points of the human body trunk of the person successfully matched with the cigarette end target point in the visible light image, and finishing the manufacture of the sample data set. The invention saves the process of manual labeling and reduces the difficulty of manufacturing the data set.

Description

Smoking behavior sample generation method and system based on double-spectrum imaging
Technical Field
The application belongs to the field of artificial intelligence safety monitoring, and particularly relates to a smoking behavior sample generation method and system based on dual-spectrum imaging.
Background
Smoking is an important potential safety hazard in many production and construction scenes such as storage, laboratories, decoration and the like. Limited by cost, the current common detection of smoking is smoke alarms or automatic identification by visible light cameras.
At present, intelligent monitoring technology based on Convolutional Neural Network (CNN) is developed more and more, and CNN is a feedforward Neural Network based on a receptive field and including Convolutional calculation, and is one of deep learning representing algorithms. CNN completes classification or regression tasks by continuously extracting image features in conjunction with the full connectivity layer. The method is mainly used for tasks such as image classification and target detection. However, the weight parameters in the CNN network need to be trained by means of data, and the more the general data is, the better the regularization effect of the trained parameters is. While there are many published datasets on the web today, these are common and not specific to a particular task. In the smoking detection task, a large number of images of people smoking need to be collected, and at present, no richer public data set related to the task exists, so that manual collection is needed. The collected data needs manual marking, induction, arrangement and other procedures, and the method is a very difficult task.
Disclosure of Invention
The application aims to provide a smoking behavior sample generation method and system based on dual-spectrum imaging, so that the manual labeling process is omitted, and the data set manufacturing difficulty is reduced.
In order to achieve the purpose, the technical scheme adopted by the application is as follows:
a method of dual spectral imaging-based smoking behavior sample generation, the dual spectral imaging including imaging by a visible light camera and an infrared imaging camera, the method of dual spectral imaging-based smoking behavior sample generation comprising:
step 1, performing pixel matching on a visible light camera and an infrared imaging camera to obtain a homography transformation matrix;
step 2, acquiring a visible light image based on the visible light camera, and carrying out single skeleton detection on the people in the visible light image to obtain the key points of the human trunk of each person;
step 3, acquiring an infrared image based on the infrared imaging camera, carrying out high-temperature point detection on the infrared image, determining a cigarette end target point in the infrared image, and calculating pixel coordinates of the cigarette end target point in the visible light image through a homography transformation matrix;
step 4, matching the figure and the cigarette end target point, comprising the following steps:
step 4.1, calculating the matching probability of the cigarette end target point and the human body trunk key point:
Figure BDA0002869679490000021
in the formula, PtijRepresenting the matching probability of the human body trunk key point t of the character i and the cigarette end target point j, wherein mu and sigma are matching coefficients, and x is the pixel distance between the cigarette end target point j and the human body trunk key point t of the character i;
step 4.2, if the matching probability of the cigarette end target point and the human body trunk key point is greater than the matching threshold value, establishing a connection relation between the corresponding cigarette end target point and the human body trunk key point of the character;
4.3, calculating the non-matching probability of the cigarette end target point:
Figure BDA0002869679490000022
in the formula, PjIs the no-match probability of the cigarette end target point j, k is the number of people having a connection relation with the cigarette end target point j, and m belongs to [1, k ∈]，PmThe maximum matching probability between the mth figure and the cigarette end target point j is obtained;
step 4.4, determining the optimal matching of the cigarette end target point:
Figure BDA0002869679490000023
in the formula, τjThe best match is made for the cigarette end target point j, n is the total number of the cigarette end target points, j belongs to [1, n ∈]，P′jNon-matching probability P containing cigarette end target point jjAnd the matching probability P with the key points of the human body trunktij；
And 5, marking the minimum outer envelope rectangle of the human body key points of the person successfully matched with the cigarette end target point in the visible light image, and finishing the manufacture of the sample data set.
Several alternatives are provided below, but not as an additional limitation to the above general solution, but merely as a further addition or preference, each alternative being combinable individually for the above general solution or among several alternatives without technical or logical contradictions.
Preferably, the pixel matching between the visible light camera and the infrared imaging camera includes:
adopting an 8 multiplied by 8 checkerboard calibration board to record the positions of corner pixels in a visible light image acquired by a visible light camera
Figure BDA0002869679490000024
And recording in infrared imagingPosition omega { (u) in infrared image acquired by camerai，vi)，i＝1，2，3...49}；
Calculating position
Figure BDA0002869679490000031
And the homography transformation matrix H between the two positions ω, such that
Figure BDA0002869679490000032
And completing pixel matching.
Preferably, the human trunk key points comprise head key points h of a skeletons(u, v), left wrist Key Point ls(u, v), right wrist Key Point rs(u, v) while recording the skeleton head width l.
Preferably, the matching coefficients μ and σ include:
if the cigarette end target point is matched with the head key point of the skeleton, then
Figure BDA0002869679490000033
If the cigarette end target point is matched with the left wrist key point or the right wrist key point, mu is l,
Figure BDA0002869679490000034
preferably, the detecting the high temperature point of the infrared image and determining the target point of the cigarette end in the infrared image includes:
marking pixel points of the infrared image with the temperature larger than the temperature threshold value to obtain one or more suspicious regions;
filtering suspicious regions with the number of pixel points smaller than a number threshold value in the suspicious regions;
and taking the pixel point with the highest temperature in each remaining suspicious region as a cigarette end target point.
Preferably, the method for generating smoking behavior samples based on dual-spectrum imaging further comprises performing data enhancement on the visible light image after the human skeleton envelope is marked.
The application also provides a system for generating the smoking behavior sample based on dual-spectrum imaging, wherein the system for generating the smoking behavior sample based on dual-spectrum imaging comprises a visible light camera, an infrared imaging camera and a main controller, and the main controller is respectively connected with the visible light camera and the infrared imaging camera;
the master controller comprises a memory having stored therein a computer program and a processor executing the computer program to implement the steps of the method of dual spectral imaging-based smoking behavior sample generation.
The application provides a smoking behavior sample generation's method and system based on two spectral imaging, cooperation through visible light camera and infrared imaging camera obtains personage information and cigarette end information, and through the preparation to smoking pedestrian's data set is accomplished in matching between them, the process that needs the artificial a large amount of smoking pedestrian's images of searching for has been removed from, the process that needs the artificial image of carrying out the image annotation has also been removed from with the mark through automatic matching simultaneously, smoking behavior data set manufacture process has greatly been simplified, and be favorable to obtaining that the data bulk is bigger, the more abundant data set of image.
Drawings
FIG. 1 is a flow chart of a method of dual spectral imaging-based smoking behavior sample generation according to the present application;
FIG. 2 is a schematic diagram of the present application illustrating the connection between the cigarette butt target point and the human torso key point of the character;
fig. 3 is a schematic structural diagram of a system for smoking behavior sample generation based on dual spectral imaging according to the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the present application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.
In one embodiment, a smoking behavior sample generation method based on dual-spectrum imaging is provided, wherein the dual-spectrum imaging refers to visible light imaging and infrared imaging, namely the method can be realized based on a visible light camera and an infrared imaging camera, and the acquisition and labeling of smoking behavior images are automatically completed.
As shown in fig. 1, the method for generating smoking behavior sample based on dual spectral imaging of the embodiment includes the following steps:
step 1, performing pixel matching on a visible light camera and an infrared imaging camera to obtain a homography transformation matrix. The homography transformation matrix is calculated to establish a corresponding relation between the collected visible light image and the infrared image so as to perform comprehensive analysis on the visible light image and the infrared image at the same time and in the same scene. The pixel matching is a mature technology in the field of image processing, and the embodiment does not specifically limit the manner used, and may be implemented, for example, in the following manner:
adopting an 8 multiplied by 8 checkerboard calibration board to record the positions of corner pixels in a visible light image acquired by a visible light camera
Figure BDA0002869679490000041
And a position ω { (u) } recorded in the infrared image acquired by the infrared imaging camerai，vi) I 1, 2, 3.. 49}, and then calculating the position
Figure BDA0002869679490000042
And the homography transformation matrix H between the two positions ω, such that
Figure BDA0002869679490000043
And completing pixel matching. Wherein the computation homography transformation matrix can be computed using the function findHomography in opencv.
And 2, acquiring a visible light image based on the visible light camera, and carrying out single skeleton detection on the people in the visible light image to obtain the key points of the human trunk of each person.
From the perspective of executing the steps, there is no strict sequence between processing the visible light image and the infrared image, but the present embodiment suggests processing the visible light image first considering that the accuracy of identifying whether a person exists in the scene is higher than the accuracy of identifying whether a smoke head high temperature point exists in the scene.
Namely, when the visible light image is subjected to single skeleton detection, people do not exist in the current scene, the subsequent steps are not executed continuously, and the execution is finished, so that the next image acquisition and processing is continued, and the data set manufacturing speed is improved.
Since the key points of the human body trunk involved by the smoking pedestrian are mainly at the head and the wrist, the key points of the human body trunk in the embodiment include the key point h of the head of the skeletons(u, v), left wrist Key Point ls(u, v), right wrist Key Point rs(u, v) and recording the width l (pixel unit) of the skeleton head, wherein the width of the skeleton head can be obtained according to the outline of the face.
It should be noted that, extracting the human body trunk point is a mature technology in the field of human recognition, and a specific extraction method is not limited in this embodiment, and may be, for example, a GCN network, an openpos network, and the like, and can obtain the key point and the face contour, and obtain the head width by calculating the width of the contour envelope.
And 3, acquiring an infrared image based on the infrared imaging camera, carrying out high-temperature point detection on the infrared image, determining a cigarette end target point in the infrared image, and calculating pixel coordinates of the cigarette end target point in the visible light image through a homography transformation matrix.
The infrared image can present pixel points with different temperatures, the cigarette end target point is determined from the infrared image based on the characteristic, and each continuous high-temperature pixel point area can be used as one cigarette end target point when the cigarette end target point is determined. However, in order to reduce environmental interference, in one embodiment, the method of determining the plume target point is as follows:
marking pixel points with the temperature higher than a temperature threshold (for example, 150 ℃) in the infrared image to obtain one or more suspicious regions; filtering out suspicious regions with the number of pixel points smaller than a number threshold (for example, 5) in the suspicious regions; and taking the pixel point with the highest temperature in each remaining suspicious region as a cigarette end target point.
It should be noted that if no suspicious region exists after filtering, it indicates that no smoking behavior exists in the current environment, the visible light image and the infrared image acquired this time are discarded, and the step 2 is returned again to perform the data set creation for a new time.
Step 4, associating the character and the cigarette end target point, comprising the following steps:
step 4.1, calculating the matching probability of the cigarette end target point and the human body trunk key point:
Figure BDA0002869679490000051
in the formula, PtijRepresenting the matching probability of the human body trunk key point t of the character i and the cigarette end target point j, i belongs to [1, q ]]Q is the number of people in the visible light image obtained at this time, and j belongs to [1, n ]]N is the total number of cigarette end target points, and t belongs to [1, r ]]R number of types of human trunk key points, where r is 3 in this embodiment, μ and σ are matching coefficients, and x is a pixel distance between the cigarette end target point j and the human trunk key point t of the person i.
Because there is a difference in the probability of the butt target point at the head and the wrist, in one embodiment, if the butt target point matches the head key point of the skeleton, then
Figure BDA0002869679490000061
If the cigarette end target point and the left wrist key point or the right wrist key pointAnd matching is carried out, then mu is equal to l,
Figure BDA0002869679490000062
the accuracy of the calculation of the matching probability of the cigarette end target point and the human body trunk key point is improved.
And 4.2, if the matching probability of the cigarette end target point and the human body trunk key point is greater than the matching threshold (for example, 0.4), establishing a connection relation (namely, connecting lines, as shown in fig. 2) between the corresponding cigarette end target point and the human body trunk key point of the person.
4.3, calculating the non-matching probability of the cigarette end target point:
Figure BDA0002869679490000063
in the formula, PjIs the no-match probability of the cigarette end target point j, k is the number of people having a connection relation with the cigarette end target point j, and m belongs to [1, k ∈]，PmThe maximum matching probability between the mth person and the cigarette end target point j. The non-matching probability considers that the current cigarette end target point may be an interference point or other situations, that is, when the embodiment finds the best matching of each cigarette end target point, the matching between the cigarette end target point and the person is considered, and meanwhile, the situation that the person is not matched is also considered, so that the situation that the cigarette end target point and the person are forcibly matched is avoided, and the reliability of the finally obtained data set is improved.
Step 4.4, determining the optimal matching of the cigarette end target point:
Figure BDA0002869679490000064
in the formula, τjIs the best match, P ', to the butt target point j'jNon-matching probability P containing cigarette end target point jjAnd the matching probability P with the key points of the human body trunktij. Namely, the matching condition with the highest probability is selected from the matching probability and the non-matching probability of the cigarette end target point and the human body trunk key point as the cigarette end target point.
And 4, marking the minimum outer envelope rectangle of the human body key points of the person successfully matched with the cigarette end target point in the visible light image, and finishing the manufacture of the sample data set.
In the embodiment, when labeling is performed, it is preferable to label the pixel coordinates of the four end corner positions of the minimum rectangular frame corresponding to the human skeleton of the figure, so as to accurately record the position of the smoking figure in the image, and store the labeling information in an xml file.
In practical application, the steps 2-4 are repeated continuously to obtain enough samples, and image acquisition can be performed on the basis of multiple pairs of visible light cameras and infrared imaging cameras installed in different scenes, so that the richness of a sample set is guaranteed.
In addition, the method for generating the smoking behavior sample based on the dual-spectrum imaging of the embodiment further comprises the step of performing data enhancement on the visible light image after the human skeleton envelope is marked so as to expand the data set.
And performing data enhancement on the marked visible light image, for example, performing 7 modes of rotating 90 degrees, rotating 180 degrees, rotating 270 degrees, horizontally turning, vertically turning, rotating 90 degrees and horizontally turning, and rotating 90 degrees and vertically turning, wherein the data volume of the whole data set is increased by 7 times.
In another embodiment, as shown in fig. 3, there is also provided a dual spectral imaging based smoking behaviour sample generation system, which comprises a visible light camera 2, an infrared imaging camera 1 and a main controller 3.
The main controller 3 in this embodiment is connected to the visible light camera 2 and the infrared imaging camera 1 (network cable connection), and may be a video processing server, and includes a memory and a processor, where the memory stores a computer program, and the processor executes the computer program to implement the steps of the method for generating smoking behavior samples based on dual-spectrum imaging.
The master controller may be a terminal including a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the master controller is configured to provide computing and control capabilities. The memory of the main controller includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the main controller is used for connecting and communicating with an external terminal through a network. The display screen of the main controller can be a liquid crystal display screen or an electronic ink display screen, and the input device of the main controller can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the main controller, an external keyboard, a touch pad or a mouse and the like
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (7)

1. A method for generating smoking behavior samples based on dual-spectrum imaging, wherein the dual-spectrum imaging comprises imaging of a visible light camera and an infrared imaging camera, and the method for generating smoking behavior samples based on the dual-spectrum imaging comprises the following steps:
step 1, performing pixel matching on a visible light camera and an infrared imaging camera to obtain a homography transformation matrix;
step 2, acquiring a visible light image based on the visible light camera, and carrying out single skeleton detection on the people in the visible light image to obtain the key points of the human trunk of each person;
step 3, acquiring an infrared image based on the infrared imaging camera, carrying out high-temperature point detection on the infrared image, determining a cigarette end target point in the infrared image, and calculating pixel coordinates of the cigarette end target point in the visible light image through a homography transformation matrix;
step 4, matching the figure and the cigarette end target point, comprising the following steps:
step 4.1, calculating the matching probability of the cigarette end target point and the human body trunk key point:
Figure FDA0002869679480000011
in the formula, PtijRepresenting the matching probability of the human body trunk key point t of the character i and the cigarette end target point j, wherein mu and sigma are matching coefficients, and x is the pixel distance between the cigarette end target point j and the human body trunk key point t of the character i;
step 4.2, if the matching probability of the cigarette end target point and the human body trunk key point is greater than the matching threshold value, establishing a connection relation between the corresponding cigarette end target point and the human body trunk key point of the character;
4.3, calculating the non-matching probability of the cigarette end target point:
Figure FDA0002869679480000012
in the formula, PjIs the no-match probability of the cigarette end target point j, k is the number of people having a connection relation with the cigarette end target point j, and m belongs to [1, k ∈]，PmThe maximum matching probability between the mth figure and the cigarette end target point j is obtained;
step 4.4, determining the optimal matching of the cigarette end target point:
Figure FDA0002869679480000013
in the formula, τjThe best match is made for the cigarette end target point j, n is the total number of the cigarette end target points, j belongs to [1, n ∈]，P′jNon-matching probability P containing cigarette end target point jjAnd the matching probability P with the key points of the human body trunktij；
And 5, marking the minimum outer envelope rectangle of the human body key points of the person successfully matched with the cigarette end target point in the visible light image, and finishing the manufacture of the sample data set.
2. The method of dual spectral imaging-based smoking behavior sample generation according to claim 1, wherein pixel matching a visible light camera with an infrared imaging camera comprises:
adopting an 8 multiplied by 8 checkerboard calibration board to record the positions of corner pixels in a visible light image acquired by a visible light camera
Figure FDA0002869679480000021
And a position ω { (u) } recorded in the infrared image acquired by the infrared imaging camerai，vi)，i＝1，2，3...49}；
Calculating position
Figure FDA0002869679480000022
And the homography transformation matrix H between the two positions ω, such that
Figure FDA0002869679480000023
And completing pixel matching.
3. The method of dual spectral imaging-based smoking behavior sample generation according to claim 1, wherein the human torso keypoints comprise skeletal head keypoints hs(u, v), left wrist Key Point ls(u, v), right wrist Key Point rs(u, v) while recording the skeleton head width l.
4. The method of dual spectral imaging-based smoking behavior sample generation according to claim 3, wherein the matching coefficients μ, σ, comprise:
if the cigarette end target point is matched with the head key point of the skeleton, then
Figure FDA0002869679480000024
If the cigarette end target point is matched with the left wrist key point or the right wrist key point, mu is l,
Figure FDA0002869679480000025
5. the method of dual spectral imaging-based smoking behavior sample generation according to claim 1, wherein said performing high temperature point detection on the infrared image and determining a butt target point in the infrared image comprises:
marking pixel points of the infrared image with the temperature larger than the temperature threshold value to obtain one or more suspicious regions;
filtering suspicious regions with the number of pixel points smaller than a number threshold value in the suspicious regions;
and taking the pixel point with the highest temperature in each remaining suspicious region as a cigarette end target point.
6. The method of dual spectral imaging-based smoking behavior sample generation according to claim 1, further comprising data enhancement of the visible light image after labeling the human skeletal envelope.
7. A smoking behavior sample generation system based on dual-spectrum imaging is characterized by comprising a visible light camera, an infrared imaging camera and a main controller, wherein the main controller is respectively connected with the visible light camera and the infrared imaging camera;
the master controller comprises a memory having stored therein a computer program and a processor executing the computer program to implement the steps of the method of dual spectral imaging based smoking behaviour sample generation according to any one of claims 1 to 6.
CN202011593064.9A 2020-12-29 2020-12-29 Smoking behavior sample generation method and system based on double-spectrum imaging Active CN112651347B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011593064.9A CN112651347B (en) 2020-12-29 2020-12-29 Smoking behavior sample generation method and system based on double-spectrum imaging

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011593064.9A CN112651347B (en) 2020-12-29 2020-12-29 Smoking behavior sample generation method and system based on double-spectrum imaging

Publications (2)

Publication Number Publication Date
CN112651347A true CN112651347A (en) 2021-04-13
CN112651347B CN112651347B (en) 2022-07-05

Family

ID=75363913

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011593064.9A Active CN112651347B (en) 2020-12-29 2020-12-29 Smoking behavior sample generation method and system based on double-spectrum imaging

Country Status (1)

Country Link
CN (1) CN112651347B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105913040A (en) * 2016-04-27 2016-08-31 重庆邮电大学 Real time double cameras provided pedestrian detection system for use under scotopic vision conditions
CN108759844A (en) * 2018-06-07 2018-11-06 科沃斯商用机器人有限公司 Robot relocates and environmental map construction method, robot and storage medium
CN109804622A (en) * 2016-09-30 2019-05-24 微软技术许可有限责任公司 Infrared image stream is restained
CN110081982A (en) * 2019-03-11 2019-08-02 中林信达（北京）科技信息有限责任公司 A kind of unmanned plane target localization method based on double spectrum photoelectric search
JP6617237B1 (en) * 2019-04-03 2019-12-11 株式会社Ｍｕｊｉｎ Robot system, robot system method and non-transitory computer readable medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105913040A (en) * 2016-04-27 2016-08-31 重庆邮电大学 Real time double cameras provided pedestrian detection system for use under scotopic vision conditions
CN109804622A (en) * 2016-09-30 2019-05-24 微软技术许可有限责任公司 Infrared image stream is restained
CN108759844A (en) * 2018-06-07 2018-11-06 科沃斯商用机器人有限公司 Robot relocates and environmental map construction method, robot and storage medium
CN110081982A (en) * 2019-03-11 2019-08-02 中林信达（北京）科技信息有限责任公司 A kind of unmanned plane target localization method based on double spectrum photoelectric search
JP6617237B1 (en) * 2019-04-03 2019-12-11 株式会社Ｍｕｊｉｎ Robot system, robot system method and non-transitory computer readable medium

Also Published As

Publication number Publication date
CN112651347B (en) 2022-07-05

Similar Documents

Publication Publication Date Title
Zhang et al. Co-saliency detection via mask-guided fully convolutional networks with multi-scale label smoothing
Li et al. Contour knowledge transfer for salient object detection
CN108764065B (en) Pedestrian re-recognition feature fusion aided learning method
CN107491726B (en) Real-time expression recognition method based on multichannel parallel convolutional neural network
CN111814661B (en) Human body behavior recognition method based on residual error-circulating neural network
Bhattacharya et al. Recognition of complex events: Exploiting temporal dynamics between underlying concepts
Hsu et al. Unsupervised cnn-based co-saliency detection with graphical optimization
Hou et al. Content-attention representation by factorized action-scene network for action recognition
CN107145862B (en) Multi-feature matching multi-target tracking method based on Hough forest
CN109145759A (en) Vehicle attribute recognition methods, device, server and storage medium
CN110555481A (en) Portrait style identification method and device and computer readable storage medium
CN109766840A (en) Facial expression recognizing method, device, terminal and storage medium
CN111783576A (en) Pedestrian re-identification method based on improved YOLOv3 network and feature fusion
CN109086659B (en) Human behavior recognition method and device based on multi-channel feature fusion
CN110222572A (en) Tracking, device, electronic equipment and storage medium
CN113627402B (en) Image identification method and related device
CN109670517A (en) Object detection method, device, electronic equipment and target detection model
CN114821014A (en) Multi-mode and counterstudy-based multi-task target detection and identification method and device
CN113378675A (en) Face recognition method for simultaneous detection and feature extraction
Jain et al. Literature review of vision‐based dynamic gesture recognition using deep learning techniques
Ma et al. Loop closure detection via locality preserving matching with global consensus
CN112651347B (en) Smoking behavior sample generation method and system based on double-spectrum imaging
CN110969173A (en) Target classification method and device
CN113705301A (en) Image processing method and device
CN112712051A (en) Object tracking method and device, computer equipment and storage medium

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
GR01 Patent grant
GR01 Patent grant