CN205184784U - Machine people goes on patrol - Google Patents
Machine people goes on patrol Download PDFInfo
- Publication number
- CN205184784U CN205184784U CN201520963772.5U CN201520963772U CN205184784U CN 205184784 U CN205184784 U CN 205184784U CN 201520963772 U CN201520963772 U CN 201520963772U CN 205184784 U CN205184784 U CN 205184784U
- Authority
- CN
- China
- Prior art keywords
- digital picture
- image
- pedestrian
- patrol robot
- 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.)
- Active
Links
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The utility model discloses a machine people goes on patrol. This machine people goes on patrol marchs the in -process and acquires the real time simulation image, will real time simulation image conversion is digital image to make statistics of out the number among the digital image through intelligent analysis. Be greater than under the condition of setting for the threshold value in the number that the judgement was makeed statistics of out, machine people goes on patrol halts to send alarm information, current GPS location information for remote monitering system. According to the utility model discloses embodiment can send the early warning under the circumstances of discovery personnel gathering, remind the security personnel to pay close attention to.
Description
Technical field
The utility model relates to intelligent robot technology field, is specifically related to a kind of patrol robot.
Background technology
Along with socioeconomic development, in living community, supermarket, conference and exhibition center, station, the place such as airport flow of the people increasing, the increase of activity number easily causes various accident to occur.Now traditional fixing camera monitoring can not effectively frighten and prevent crime, cannot meet and day by day increase security demand.Therefore, mobile video monitor patrol robot arises at the historic moment.
At present, the mobile patrol robot on market has camera, and can provide video record memory function.But this simple video record and memory function cannot realize finding fast or early warning the accident in patrol monitoring range.
Utility model content
In view of this, the utility model embodiment provides a kind of patrol robot and method for supervising thereof, solves the early warning problem to accident.
According to embodiment of the present utility model, provide a kind of patrol robot, it comprises camera, memory, GPS module and processor.Described camera is configured to, and obtains real-time Simulation image, and described real-time Simulation image is converted to digital picture.Described memory is configured to, and described digital picture is stored with local high definition form.Described GPS module is configured to, and obtains the positional information of described patrol robot.Described processor is configured to, from memory, digital picture described in a width is extracted according to some cycles, employing intelligent analysis method counts the number in a described width digital picture, and judge whether the number of adding up is greater than setting threshold value, when added up number is greater than setting threshold value, control described patrol robot to halt, and warning message and current location information are sent to remote monitoring end.
Further, described processor is also configured to, and receives the described digital picture that camera is converted to, described digital image compression is encoded, and send to described remote monitoring end.
Further, described setting threshold value was arranged according to time and/or GPS information.
Further, count the number in a described width digital picture, comprise described in described employing intelligent analysis method, employing pedestrian detection method counts the pedestrian's quantity in a width digital picture; And/or employing method for detecting human face counts the face quantity in a width digital picture.
Further, described pedestrian and/or described face can be gone out by frame at described remote monitoring end.
Further, described setting threshold value was arranged according to time and/or GPS information.
Further, the number counted in a described width digital picture by described pedestrian detection method is comprised: digital picture described in a width is divided into the identical subgraph of several sizes; The histograms of oriented gradients feature extracting the identical subgraph of several sizes described is respectively input to pedestrian detection model; Counted the number in digital picture described in a width by the judgement of described pedestrian detection model, namely determine the position of pedestrian in image, thus count the quantity of pedestrian in image.
Further, the training process of described pedestrian detection model comprises: set up sample, and wherein select pedestrian's image as positive sample, non-pedestrian image is as negative sample; Extract the histograms of oriented gradients feature of described positive sample, described negative sample respectively; The described histograms of oriented gradients feature adopting SVM algorithm to align sample and negative sample is carried out training and is obtained pedestrian detection model.
Further, described employing method for detecting human face counts the face quantity in a width digital picture, specifically comprises: a width digital picture is divided into the identical subgraph of several sizes; The Face datection model that described in extracting respectively, the Haar feature of the subgraph that several sizes are identical trains before being input to; Gone out the face in digital picture described in a width by described Face datection model inspection, namely determine the position of face in image, thus count the quantity of face in image.
Further, the training process of described Face datection model comprises: set up sample, and wherein select facial image as positive sample, non-face image is as negative sample; Extract the Haar feature of described positive sample, described negative sample respectively; The described Haar feature adopting Adaboost algorithm to align sample and negative sample is carried out training and is obtained Face datection model.
Further, a kind of patrol robot is provided, comprises: camera, for obtaining real-time Simulation image, and by described real-time Simulation image uploading to video frequency collection card; Video frequency collection card, converts digital picture to for the described real-time Simulation image transmitted by camera, and described digital picture is sent to main frame; GPS module, for obtaining the positional information of described patrol robot; Driver module, moves for driving described patrol robot; And main frame, for being counted the number in described digital picture by intellectual analysis, and judge whether the number counted is greater than setting threshold value, when added up number is greater than setting threshold value, control described patrol robot by described driver module to halt, and obtained the current location information of described patrol robot by described GPS module, and warning message and current location information are sent to remote monitoring end.
The utility model embodiment provides a kind of patrol robot, and it can carry out intellectual analysis to realtime graphic, and adds up the number in realtime graphic, i.e. the quantity of pedestrian and face.When the quantity of pedestrian and face is greater than setting threshold value, sends early warning, and send current location information to remote monitoring end, remind Security Personnel to pay close attention to.
Accompanying drawing explanation
Figure 1 shows that the flow chart of the patrol robot execution method for supervising that the utility model one embodiment provides;
Figure 2 shows that the flow chart of the pedestrian's quantity in the width digital picture that employing pedestrian detection method that the utility model one embodiment provides is added up each second and taken out;
Figure 3 shows that the training process flow chart of the pedestrian detection model that the utility model one embodiment provides;
Figure 4 shows that the flow chart of the face quantity in the width digital picture that employing method for detecting human face that the utility model one embodiment provides is added up each second and extracted;
Figure 5 shows that the training process flow chart of the Face datection model that the utility model one embodiment provides;
Figure 6 shows that the patrol robot system architecture diagram that the utility model one embodiment provides.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the utility model embodiment, be clearly and completely described the technical scheme in the utility model embodiment, obviously, described embodiment is only the utility model part embodiment, instead of whole embodiments.Based on the embodiment in the utility model, those of ordinary skill in the art are not making the every other embodiment obtained under creative work prerequisite, all belong to the scope of the utility model protection.
According to embodiment of the present utility model, provide a kind of patrol robot, it comprises camera, memory, GPS module and processor.Fig. 1 illustrates the flow chart of the patrol robot execution method for supervising that the utility model one embodiment provides.Specifically comprise:
S101: camera obtains real-time Simulation image.
S102: real-time Simulation image is converted to digital picture by camera.
S103: digital picture is stored in memory with local high definition form.
S104: processor extracts digital picture described in a width according to some cycles, employing intelligent analysis method counts the number in digital picture described in a width.Preferably, a width digital picture is extracted each second.
S105: processor judges whether number is greater than setting threshold value, if be greater than, robot halts, and the current GPS location information that warning message and GPS module obtain is sent to remote monitoring end.
Patrol robot according to the present embodiment can go out number in monitor video by Intelligent statistical, when the number of adding up exceeds setting threshold value, sends early warning information to remote monitoring end, has the prevention effect well to accident.
According to an embodiment of the present utility model, processor is also configured to, and receives the described digital picture that camera is converted to, described digital image compression is encoded, and send to described remote monitoring end.Realize the real-time display of video image at remote monitoring end, be convenient to the early warning information of Security Personnel to robot on the one hand and confirm, check and whether accident occurs; Make monitor procedure more transparent on the other hand.
According to an embodiment of the present utility model, employing intelligent analysis method counts the number in a described width digital picture, comprising: employing pedestrian detection method counts the pedestrian's quantity in a width digital picture; And/or employing method for detecting human face counts the face quantity in a width digital picture.Especially, when pedestrian is back to camera time, can pedestrian be detected, but can not face be detected; And when face occupies most of space of picture, face can be detected, but pedestrian cannot be detected; In view of the foregoing, the function simultaneously with pedestrian detection and Face datection is just needed.
According to an embodiment of the present utility model, the digital picture that remote monitoring termination is sent after receiving treated device compressed encoding, shows after decoding over the display, and the pedestrian in the image of display display can be gone out by frame.In this way, can give top priority to what is the most important, accelerate the target search time of monitor staff.
According to an embodiment of the present utility model, setting threshold value can according to time and GPS information classification setting.Find 5 people when such as can set daytime and send warning message above, when evening, finding that namely 1 people sends warning message.Specific to the time, one day 24 hours can certainly be divided into several intervals, according to the habits and customs reasonable set alarm threshold value of people.Can setting for another example when finding more than 1 people to send warning message near enclosure wall, finding that more than 10 people send warning message on square or doorway.It should be known that classification setting standard is here only exemplary.In like manner, face quantity occurs that abnormal alarm mode is identical therewith.
Figure 2 shows that the flow chart of the pedestrian's quantity in the width digital picture that employing pedestrian detection method that the utility model one embodiment provides is added up each second and taken out, specifically comprise:
S201: a width digital picture is divided into the identical subgraph of several sizes.Preferably, adopt the mode of the sliding window of rectangle from described digital picture, obtain the identical subgraph of some sizes.
S202: the histograms of oriented gradients feature extracting the identical subgraph of several sizes is input to pedestrian detection model.In piece image, the presentation of localized target and shape can be described well by the direction Density Distribution at gradient or edge, and the gradient of each pixel can be used for describing its direction density, and the gradient orientation histogram of image therefore can be utilized to state the feature of image.
S203: count the pedestrian's quantity in a width digital picture by the judgement of pedestrian detection model.Namely determine the position of pedestrian in image, thus count the quantity of pedestrian in image.
Specifically describe the process that a width digital picture carries out pedestrian detection below.First, extract a width digital picture and it can be used as current goal digital picture, set up the sliding window of rectangle that size is 50*20, from left to right slide until image low order end from the upper left corner of figure according to certain displacement, then sliding window moves down certain displacement, again scan from left to right, until travel through entire image completely; Rectangular area is moved the rear portion intercepts surrounded out as subgraph at every turn, and extract the histograms of oriented gradients feature of this subgraph; The histograms of oriented gradients feature of subgraph is inputed to pedestrian detection model and carries out pedestrian's differentiation.Keep the size of the sliding window of rectangle constant, according to certain scale smaller target digital image, form new target digital image, according to the target digital image that said process traverse scanning is new, and extract the histograms of oriented gradients feature of all subgraphs, give pedestrian detection model respectively and carry out pedestrian's differentiation.Repeat said process, until have one to slide the width or highly equal of window with rectangle in the width of target image or height.The position of pedestrian in image can be detected by said process, thus determine the quantity of pedestrian further.
In step S203, Figure 3 shows that the training process flow chart of the pedestrian detection model that the utility model one embodiment provides, it comprises:
S301: set up sample, wherein select pedestrian's image as positive sample, non-pedestrian image is as negative sample.Such as, first, find some to include the image of pedestrian, in cut-away view picture, comprise the rectangular area of pedestrian, be normalized into the size of 50*20, form positive sample; Look for some not comprise the image of pedestrian, some rectangular areas of random intercepting, are normalized into the size of 50*20, form negative sample; Indicate that all samples are positive sample or negative sample, be the label that each sample sticks a positive negative sample.
S302: the histograms of oriented gradients feature extracting described positive sample, described negative sample respectively.
S303: adopt SVMs (SVM; Supportvectormachine) the described histograms of oriented gradients feature that algorithm aligns sample and negative sample is carried out training and is obtained pedestrian detection model.SVM is a kind of machine learning method, and the data of giving SVM had both comprised the histograms of oriented gradients feature of positive sample and negative sample, also comprises label corresponding with it.
Figure 4 shows that the flow chart of the face quantity in the width digital picture that employing method for detecting human face that the utility model one embodiment provides is added up each second and extracted, specifically comprise:
S401: a width digital picture is divided into the identical subgraph of some sizes.Use the mode of the sliding window of rectangle from described digital picture, obtain the identical subgraph of some sizes.
S402: the Face datection model trained before the Haar feature extracting the identical subgraph of several sizes is respectively input to.Haar feature, is divided into four classes: edge feature, linear character, central feature and diagonal feature, be combined into feature templates.Adularescent and black two kinds of rectangles in feature templates, and the characteristic value defining this template be white rectangle pixel and deduct black rectangle pixel and.Haar characteristic value reflects the grey scale change situation of image.Such as: some features of face simply can be described by rectangular characteristic, as: eyes are darker than cheek color, and bridge of the nose both sides are darker than bridge of the nose color, and face is darker etc. than ambient color.Therefore Haar feature can as feature statement required during Face datection.
S403: go out the face quantity in a width digital picture by Face datection model inspection.Namely determine the position of face in image, and then the quantity of face in piece image can be counted.
Specifically describe the process that a width digital picture carries out Face datection below.First, extract a width digital picture and it can be used as current goal digital picture, set up the sliding window of rectangle that size is 24*24, from left to right slide until image low order end from the upper left corner of figure according to certain displacement, then sliding window moves down certain displacement, again scan from left to right, until travel through entire image completely; Rectangular area is moved the rear portion intercepts surrounded out as subgraph at every turn, and extract the Haar feature of this subgraph; The Haar feature of subgraph is inputed to Face datection model and carries out human face discriminating.Keep the size of the sliding window of rectangle constant, according to certain scale smaller target digital image, form new target digital image, according to the target digital image that previous step traverse scanning is new, and extract the Haar feature of all subgraphs, give Face datection model respectively and carry out human face discriminating.Repeat said process, until the wide senior middle school of target image have one with rectangle cunning window wide height equal.The position of face in image can be detected by said process, thus determine the quantity of face further.
In step S403, Figure 5 shows that the training process flow chart of the Face datection model that the utility model one embodiment provides, it comprises:
S501: set up sample, wherein facial image is as positive sample, and non-face image is as negative sample.Such as, first, find some to include the image of face, in cut-away view picture, comprise the rectangular area of face, be normalized into the size of 24*24, form positive sample; Look for some not comprise the image of face, some rectangular areas of random intercepting, are normalized into the size of 24*24, form negative sample; Indicate that all samples are positive sample or negative sample, be the label that each sample sticks a positive negative sample.
S502: the Haar feature extracting described positive negative sample respectively.
S503: the Haar feature adopting Adaboost cascade classifier algorithm to align sample and negative sample is carried out training and obtained Face datection model.Adaboost is a kind of iterative algorithm, and its core concept trains different graders (Weak Classifier) for same training set, then these weak classifier set got up, and forms a stronger final grader (strong classifier).Its algorithm itself realizes by changing Data distribution8, and whether it is correct according to the classification of sample each among each training set, and the accuracy rate of the general classification of last time, determines the weights of each sample.Give sub classification device by the new data set revising weights to train, finally will the grader obtained be trained finally to merge, as last Decision Classfication device at every turn.
Figure 6 shows that the patrol robot system architecture diagram that the utility model embodiment provides.As can be seen from Figure 6, patrol machine robot system according to the present embodiment comprises robotic end and remote control terminal, and wherein robotic end comprises: main frame, camera, video frequency collection card, wireless communication module, GPS module, driver module, power supply; Remote monitoring end comprises: remote monitoring computer, wireless communication module, operation bench, display terminal.
Camera, for obtaining real-time Simulation image, and by described real-time Simulation image uploading to video frequency collection card.Video frequency collection card, converts digital picture to for the described real-time Simulation image transmitted by camera, and described digital picture is uploaded to main frame.Main frame, counts the quantity of pedestrian and face in described digital picture by intellectual analysis; Judge whether the quantity of pedestrian and face is greater than setting threshold value, if be greater than, robotic end halts, and current GPS position information, warning message are sent to remote monitoring end.GPS module, for obtaining the current location information of described robotic end, and is uploaded to main frame by described current location information.Wireless communication module, for the information interaction of remote control terminal and robotic end.Driver module, under the control of main frame, realizes the oriented motion of whole robot.Power supply is the energy supply of whole robot.
Alternatively, described digital image compression coding is also sent to described remote monitoring end by main frame, and/or, described digital picture is stored with local high definition form.
Remote monitoring end comprises: remote monitoring computer, brings into operation order to robotic end, and receive realtime graphic, GPS position information that robotic end sends for sending, and warning message.Wireless communication module, for the information interaction of remote control terminal and robotic end.Display terminal, for showing the image that wireless communication module receives in real time.Operation bench, send for operating personnel bring into operation, instruction out of service.
According to an embodiment of the present utility model, patrol robot is implemented to a convertible car shape, and car is provided with robot body, wherein: main frame is arranged on the centre of convertible car, and is positioned at the below of robot body; Driver module is implemented as the wheel of convertible car; Camera is arranged to the eyes of robot body; Wireless communication module and GPS module are disposed in parallel in the head rear of robot body up and down.
According to the patrol robot of the utility model embodiment, in Intelligent statistical pedestrian quantity while automatic patrol, and can provide abnormal alarm, community policing be monitored and becomes more efficient.
The foregoing is only preferred embodiment of the present utility model, not in order to limit the utility model, all within spirit of the present utility model and principle, any amendment done, equivalent replacement etc., all should be included within protection domain of the present utility model.
Claims (10)
1. a patrol robot, is characterized in that, comprises camera, memory, GPS module and processor,
Described camera is configured to, and obtains real-time Simulation image, and described real-time Simulation image is converted to digital picture;
Described memory is configured to, and described digital picture is stored with local high definition form;
Described GPS module is configured to, and obtains the positional information of described patrol robot;
Described processor is configured to, from memory, digital picture described in a width is extracted according to some cycles, employing intelligent analysis method counts the number in a described width digital picture, and judge whether the number of adding up is greater than setting threshold value, when added up number is greater than setting threshold value, control described patrol robot to halt, and warning message and current location information are sent to remote monitoring end.
2. patrol robot according to claim 1, is characterized in that, described processor is also configured to, and receives the described digital picture that camera is converted to, described digital image compression is encoded, and send to described remote monitoring end.
3. patrol robot according to claim 1 and 2, is characterized in that, described employing intelligent analysis method counts the number in a described width digital picture, comprises,
Employing pedestrian detection method counts the pedestrian's quantity in a width digital picture; And/or
Employing method for detecting human face counts the face quantity in a width digital picture.
4. patrol robot according to claim 3, is characterized in that, described pedestrian and/or described face can be gone out by frame at described remote monitoring end.
5. patrol robot according to claim 1 and 2, is characterized in that, described setting threshold value was arranged according to time and/or GPS information.
6. patrol robot according to claim 5, is characterized in that, described employing pedestrian detection method counts the pedestrian's quantity in a width digital picture, specifically comprises:
Digital picture described in one width is divided into the identical subgraph of several sizes;
The histograms of oriented gradients feature extracting the identical subgraph of several sizes described is respectively input to pedestrian detection model;
Gone out the pedestrian in digital picture described in a width by described pedestrian detection model inspection, namely determine the position of pedestrian in image, thus count the quantity of pedestrian in image.
7. patrol robot according to claim 6, is characterized in that, the training process of described pedestrian detection model comprises:
Set up sample, wherein select pedestrian's image as positive sample, non-pedestrian image is as negative sample;
Extract the histograms of oriented gradients feature of described positive sample, described negative sample respectively;
The described histograms of oriented gradients feature adopting SVM algorithm to align sample and negative sample is carried out training and is obtained pedestrian detection model.
8. patrol robot according to claim 5, is characterized in that, described employing method for detecting human face counts the face quantity in a width digital picture, specifically comprises:
Digital picture described in one width is divided into the identical subgraph of several sizes;
The Face datection model that described in extracting respectively, the Haar feature of the subgraph that several sizes are identical trains before being input to;
Gone out the face in digital picture described in a width by described Face datection model inspection, namely determine the position of face in image, thus count the quantity of face in image.
9. patrol robot according to claim 8, is characterized in that, the training process of described Face datection model comprises:
Set up sample, wherein select facial image as positive sample, non-face image is as negative sample;
Extract the Haar feature of described positive sample, described negative sample respectively;
The described Haar feature adopting Adaboost algorithm to align sample and negative sample is carried out training and is obtained Face datection model.
10. a patrol robot, is characterized in that, comprising:
Camera, for obtaining real-time Simulation image, and by described real-time Simulation image uploading to video frequency collection card;
Video frequency collection card, converts digital picture to for the described real-time Simulation image transmitted by camera, and described digital picture is sent to main frame;
GPS module, for obtaining the positional information of described patrol robot;
Driver module, moves for driving described patrol robot; And
Main frame, for being counted the number in described digital picture by intellectual analysis, and judge whether the number counted is greater than setting threshold value, when added up number is greater than setting threshold value, control described patrol robot by described driver module to halt, and obtained the current location information of described patrol robot by described GPS module, and warning message and current location information are sent to remote monitoring end.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201520963772.5U CN205184784U (en) | 2015-11-27 | 2015-11-27 | Machine people goes on patrol |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201520963772.5U CN205184784U (en) | 2015-11-27 | 2015-11-27 | Machine people goes on patrol |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN205184784U true CN205184784U (en) | 2016-04-27 |
Family
ID=55777814
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201520963772.5U Active CN205184784U (en) | 2015-11-27 | 2015-11-27 | Machine people goes on patrol |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN205184784U (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105291111A (en) * | 2015-11-27 | 2016-02-03 | 深圳市神州云海智能科技有限公司 | Patrol robot |
| WO2021027690A1 (en) * | 2019-08-14 | 2021-02-18 | 北京猎户星空科技有限公司 | Robot control method and apparatus |
| CN112587035A (en) * | 2020-12-08 | 2021-04-02 | 珠海市一微半导体有限公司 | Control method and system for mobile robot to recognize working scene |
| CN113112654A (en) * | 2021-03-30 | 2021-07-13 | 杭州海康威视数字技术股份有限公司 | Access control service host switching method and device |
-
2015
- 2015-11-27 CN CN201520963772.5U patent/CN205184784U/en active Active
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105291111A (en) * | 2015-11-27 | 2016-02-03 | 深圳市神州云海智能科技有限公司 | Patrol robot |
| WO2021027690A1 (en) * | 2019-08-14 | 2021-02-18 | 北京猎户星空科技有限公司 | Robot control method and apparatus |
| CN112587035A (en) * | 2020-12-08 | 2021-04-02 | 珠海市一微半导体有限公司 | Control method and system for mobile robot to recognize working scene |
| CN113112654A (en) * | 2021-03-30 | 2021-07-13 | 杭州海康威视数字技术股份有限公司 | Access control service host switching method and device |
| CN113112654B (en) * | 2021-03-30 | 2022-10-25 | 杭州海康威视数字技术股份有限公司 | Access control service host switching method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105291111A (en) | Patrol robot | |
| US10223597B2 (en) | Method and system for calculating passenger crowdedness degree | |
| CN108216252B (en) | Subway driver vehicle-mounted driving behavior analysis method, vehicle-mounted terminal and system | |
| CN109670441B (en) | Method, system, terminal and computer readable storage medium for realizing wearing recognition of safety helmet | |
| CA3017027C (en) | System and method for training object classifier by machine learning | |
| CN205184784U (en) | Machine people goes on patrol | |
| KR102122859B1 (en) | Method for tracking multi target in traffic image-monitoring-system | |
| CN105913037A (en) | Face identification and radio frequency identification based monitoring and tracking system | |
| US20070098222A1 (en) | Scene analysis | |
| CN108376246A (en) | A kind of identification of plurality of human faces and tracking system and method | |
| CN109815936B (en) | Target object analysis method and device, computer equipment and storage medium | |
| KR102122850B1 (en) | Solution for analysis road and recognition vehicle license plate employing deep-learning | |
| CN107248306A (en) | A kind of system for adjusting public transport transport power | |
| US20210134146A1 (en) | Tracking and alerting traffic management system using iot for smart city | |
| CN104281839A (en) | Body posture identification method and device | |
| CN105100735A (en) | People intelligent monitoring management system and management method | |
| CN104463232A (en) | Density crowd counting method based on HOG characteristic and color histogram characteristic | |
| CN104159088A (en) | System and method of remote monitoring of intelligent vehicle | |
| CN115841651B (en) | Constructor intelligent monitoring system based on computer vision and deep learning | |
| Sikirić et al. | Image representations on a budget: Traffic scene classification in a restricted bandwidth scenario | |
| CN115620175A (en) | On-spot management system based on unmanned aerial vehicle is automatic to be patrolled and examined | |
| US11893827B2 (en) | Systems and methods of detecting mask usage | |
| CN105095891A (en) | Human face capturing method, device and system | |
| CN112633249A (en) | Embedded pedestrian flow detection method based on light deep learning framework | |
| CN105574499B (en) | A kind of number detection statistics method and system based on SOC |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |