CN106022235B - Missing child detection method based on human body detection - Google Patents
Missing child detection method based on human body detection Download PDFInfo
- Publication number
- CN106022235B CN106022235B CN201610314230.4A CN201610314230A CN106022235B CN 106022235 B CN106022235 B CN 106022235B CN 201610314230 A CN201610314230 A CN 201610314230A CN 106022235 B CN106022235 B CN 106022235B
- Authority
- CN
- China
- Prior art keywords
- target
- detection
- infrared sensor
- search window
- image
- 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
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
Abstract
The invention relates to a missing child detection method based on human body detection. Under the condition of extremely low power consumption, an infrared sensor is adopted to carry out primary detection on a human body target; after the suspicious target is preliminarily detected, the camera is started again, whether the left-over child exists is confirmed by adopting a video target rechecking method, through a double-layer detection mode, the power consumption of the system can be reduced, and the false alarm rate of the detection of the left-over child can be reduced, so that the reliability of the detection of the left-over child is improved.
Description
Technical Field
The invention relates to a missing child detection method, and belongs to the technical field of video monitoring.
Background
In recent years, accidents of death of children caused by the omission of the children on school buses occur for many times, and strong social attention is attracted. In order to prevent such accidents, there is a need to technically improve the ability to detect and warn missing children in addition to the frequent safety education and supervision of the responsible persons and the children in the vehicle. With the rapid development of computer vision technology, the human body detection technology is more and more mature, and the patent "the state recognition and danger state control system for passengers staying in a vehicle" (serial numbers: CN201510779628.0, 2015) detects whether passengers leave behind a seat through a pressure sensor, and the detection method cannot distinguish people and objects, cannot detect left-behind children in corridors or other sensor-free areas, and is not high in detection precision. The document "deep Learned Attributes for crowed Scene Understanding" (CVPR, 2015) adopts a deep learning method to detect whether a human target exists in an image, but the method has a poor effect of detecting a child with a shielding condition on a vehicle. In the literature "(Automatic Head Detection for passage Flow Analysis in gas Surveillance video" (CISP, 2012), the number of people coming in and going out is counted by arranging a camera at the door, the counting precision is not high, and it cannot be accurately judged whether children are missing on the vehicle.
Disclosure of Invention
The invention particularly provides a low-power-consumption reliable detection method for missing children.
The technical scheme of the invention is as follows: under the condition of extremely low power consumption, an infrared sensor is adopted to carry out primary detection on a human body target; after the suspicious target is preliminarily detected, the camera is started again, whether the left-over child exists is confirmed by adopting a video target rechecking method, through a double-layer detection mode, the power consumption of the system can be reduced, and the false alarm rate of the detection of the left-over child can be reduced, so that the reliability of the detection of the left-over child is improved.
In order to achieve the above purpose, the invention adopts the following technical scheme, and the flow chart is shown in figure 1:
1. infrared human body detection preliminary examination
When the school bus stops, the power consumption problem of the system is particularly concerned when the storage battery of the school bus is used for supplying power to the legacy child detection equipment because the engine does not work. If the power consumption of the system is too high, the storage battery is easy to consume too much electricity, so that the school bus cannot be ignited.
The legacy child detection device of the present invention mainly comprises three modules: ARM handles board, infrared sensor, camera (take infrared LED). The ARM processing board and the infrared sensor are low in power consumption, the power consumption of the camera is high, and if the infrared LED is turned on under the condition of dark light, the power consumption is high. In order to reduce the power consumption of the system, the ARM processing board is adopted to control the power supply of each module, only the ARM processing board and the infrared sensor are powered in a silent state, and the power consumption of the two modules is very low. Only when the infrared sensor detects a suspicious object, the ARM processor is informed to power on the camera with large power consumption. The method comprises the following specific steps:
step 1: the ARM processing board inquires GPIO level connected with the infrared sensor;
step 2: if the GPIO level connected with the infrared sensor is high level, which indicates that the infrared sensor detects a target, the camera is powered on, and the power supply of the infrared sensor is turned off; otherwise, go to Step 1.
Step 3: and the ARM processing board starts a video target rechecking thread.
Step 4: and inquiring the detection result of the video target rechecking thread within a time period of 3 minutes, starting a voice alarm of the vehicle if the video target rechecking thread detects the target, starting the wireless communication module to send the image and the vehicle information in the vehicle to a school bus owner and an upper supervision department, and closing the wireless communication module and the voice module after the sending is finished.
Step 5: and turning off the camera, simultaneously electrifying the infrared sensor, and turning to Step 1.
2. Video object review
In order to reduce false alarms caused by the infrared sensor, the invention rechecks the suspicious target by adopting a video analysis method after the infrared sensor triggers the alarm. The method comprises the following specific steps:
step 1: edge calculation
Considering that gray scale difference necessarily exists between the target and the background, the invention firstly adopts gradient operators to solve the image edge on the gray scale image. Pixel point
Can be expressed as
Wherein the content of the first and second substances,f(x,y) Representing pixel points
The gray scale of the image at (a),
and
respectively represent the edges thereof
And
a gradient in direction. The specific gradient value can be represented by convolution of the image and the gradient operator template, and the Sobel operator is adopted in the invention, as shown in FIG. 2.
Pixel point
Can be expressed as
For each pixel point, if the gradient modulus m of each pixel point is greater than the threshold T1, the pixel point is considered to be an edge pixel point. In the present invention, the threshold T1 takes the empirical value of 10.
Step 2: coarse positioning of target
Performing multi-scale search on the image, specifically, setting the minimum search window size as
(i.e., the minimum size of the suspicious object, in the present invention the image size is 640 x 480,
the number of the particles is taken to be 10,
taken as 20), the maximum search window size is
(i.e., the maximum size of the suspicious object, in the present invention
The number of the samples was taken to be 60,
taken as 120). Firstly, searching is carried out, the height of a search window is unchanged, and the window width is increased by 1 after each round of searching is finished (namely, the window width is increased from the upper left corner to the lower right corner of an image) until the maximum window width is reached
(ii) a Then the height of the search window is increased by 1, and the line search is continuously repeated until the width and the height of the search window respectively reach
And
. In the multi-scale searching process, whether the following two conditions are met in each searching window is judged:
(1) the search window contains edge pixel points;
(2) there is not edge pixel on four sides of search window, also all edge pixels are inside search window.
If a certain search window meets the two conditions, the search window is considered to have a possible target, and the rectangular box is stored.
Then, all the stored rectangular frames are merged, specifically, the rectangular frames with overlapping positions are merged, the coordinates and the sizes of the starting points of the two rectangular frames after merging are the average value of the coordinates and the sizes of the two rectangular frames before merging, and meanwhile, the total number of the merged rectangular frames is recorded as the score of the rectangular frame.
Finally, for the merged rectangular box, if the score of the merged rectangular box is greater than the threshold value T2, the rectangular box is reserved, and the suspicious target is considered to be contained in the rectangular box; otherwise, the rectangular frame is deleted. In the present invention, the threshold T2 takes the empirical value of 5.
Step 3: hough circle detection
Traversing each rectangular frame detected at Step2, scanning edge pixel points in each rectangular frame, judging whether an approximate circle exists in the rectangular frame by adopting a Hough circle detection method, and recording the radius R of the circle. If T3< R < T4, then a legacy child is deemed to be present and the video object rechecks that a child object is present. If no child is left behind in all the rectangular boxes, it is considered that there is no child target in the image. Where T3 and T4 are empirical thresholds, T3=5 and T4=30 in the present invention.
The invention has the advantages that: by adopting the double-layer detection mode of infrared human body detection initial detection and video target rechecking, the power consumption of the system can be reduced, and the false alarm rate of missing child detection can be reduced.
Drawings
FIG. 1 is a flow diagram of a legacy child detection;
figure 2 Sobel operator template.
Detailed Description
A reliable detection method for missing children with low power consumption adopts an infrared sensor to carry out primary detection on a human body target; after the suspicious target is preliminarily detected, the camera is started again, whether the left-over child exists is confirmed by adopting a video target rechecking method, through a double-layer detection mode, the power consumption of the system can be reduced, and the false alarm rate of the detection of the left-over child can be reduced, so that the reliability of the detection of the left-over child is improved.
Claims (1)
1. A missing child detection method based on human body detection is characterized in that an infrared sensor is adopted to carry out primary detection on a human body target; after the suspicious target is preliminarily detected, the camera is started again, whether a left child exists is confirmed by adopting a video target rechecking method,
the process is as follows:
(1) infrared human body detection
The legacy child detection device comprises three modules: ARM handles board, infrared sensor, camera, adopts the power supply of ARM to handle the board control each module, only handles board and infrared sensor power supply for ARM under silent state, and the consumption of these two modules is all low, only when infrared sensor detects suspicious target, just notifies the ARM treater to go up the electricity for the camera, and concrete step is:
step 1: the ARM processing board inquires GPIO level connected with the infrared sensor;
step 2: if the GPIO level connected with the infrared sensor is high level, which indicates that the infrared sensor detects a target, the camera is powered on, and the power supply of the infrared sensor is turned off; otherwise, turning to Step 1;
step 3: starting a video target rechecking thread by the ARM processing board;
step 4: inquiring the detection result of the video target rechecking thread within a time period of 3 minutes, if the video target rechecking thread detects a target, starting a voice alarm of the vehicle, starting a wireless communication module to send the image and the vehicle information in the vehicle to a school bus owner and an upper supervision department, and closing the wireless communication module and the voice module after the sending is finished;
step 5: turning off the camera, simultaneously electrifying the infrared sensor, and turning to Step 1;
(2) video object review
In order to reduce false alarm caused by the infrared sensor, after the infrared sensor triggers alarm, a video analysis method is adopted to recheck a suspicious target, and the method comprises the following specific steps:
step2.1: edge calculation
Considering that gray scale difference necessarily exists between a target and a background, firstly, gradient operators are adopted on a gray scale image to obtain image edges and pixel points
Is represented by a gradient of
Wherein the content of the first and second substances,f(x,y) Representing pixel points
The gray scale of the image at (a),
and
respectively represent the edges thereof
And
the gradient of the direction, the specific gradient value is expressed by convolution of the image and the gradient operator template,
pixel point
Is expressed as
For each pixel point, if the gradient modulus value m of each pixel point is greater than the threshold value T1, the pixel point is considered as an edge pixel point, and the threshold value T1 takes the empirical value of 10;
step2.2: coarse positioning of target
Performing multi-scale search on the image, specifically, setting the minimum search window size as
I.e., the smallest size of the suspicious object, the image size is 640 x 480,
the number of the particles is taken to be 10,
taken as 20, the maximum search window size is
I.e., the maximum size of the suspicious object,
the number of the samples was taken to be 60,
taking the value as 120, searching is carried out firstly, the height of a search window is unchanged at the moment, each round of searching is finished, namely the search window reaches the lower right corner from the upper left corner of the image, then the window width is increased by 1 until the maximum window width is reached
(ii) a Then the height of the search window is increased by 1, and the line search is continuously repeated until the width and the height of the search window respectively reach
And
in the multi-scale searching process, whether the following two conditions are met in each searching window is judged:
(a) the search window contains edge pixel points;
(b) edge pixel points do not exist on the four edges of the search window, namely all the edge pixel points are in the search window,
if a certain search window meets the above two conditions, the search window is considered to have a possible target, the rectangular box is stored,
then, all the stored rectangular frames are merged, specifically, the rectangular frames with overlapping positions are merged, the coordinates and the sizes of the starting points of the two rectangular frames after merging are taken as the average value of the coordinates and the sizes of the two rectangular frames before merging, and simultaneously, the total number of the merged rectangular frames is recorded as the score of the rectangular frame,
finally, for the merged rectangular box, if the score of the merged rectangular box is greater than the threshold value T2, the rectangular box is reserved, and the suspicious target is considered to be contained in the rectangular box; otherwise, the rectangular box is deleted, and the threshold T2 takes the experience value of 5;
step2.3: hough circle detection
Traversing each detected rectangular frame of Step2.2, scanning edge pixel points in each rectangular frame, judging whether an approximate circle exists in the rectangular frame by adopting a Hough circle detection method, recording the radius R of the circle, if T3< R < T4, determining that a left child exists, and determining that a child target exists as a video target rechecking result, and if no left child exists in all the rectangular frames, determining that no child target exists in the image, wherein T3 and T4 are experience thresholds, T3=5, and T4= 30.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314230.4A CN106022235B (en) | 2016-05-13 | 2016-05-13 | Missing child detection method based on human body detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314230.4A CN106022235B (en) | 2016-05-13 | 2016-05-13 | Missing child detection method based on human body detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106022235A CN106022235A (en) | 2016-10-12 |
| CN106022235B true CN106022235B (en) | 2021-05-28 |
Family
ID=57100206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610314230.4A Active CN106022235B (en) | 2016-05-13 | 2016-05-13 | Missing child detection method based on human body detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022235B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107933476B (en) * | 2017-11-14 | 2019-08-13 | 苏州数言信息技术有限公司 | The method and system of the sensing device of the anti-forgetting of general passenger inside the vehicle |
| CN109927644A (en) * | 2017-12-15 | 2019-06-25 | 郑州宇通客车股份有限公司 | A kind of vehicle and child-resistant legacy device |
| CN108111821A (en) * | 2018-01-10 | 2018-06-01 | 深圳羚羊极速科技有限公司 | A kind of equipment for being integrally interconnected net video access gateway and edge calculations |
| US11521478B2 (en) | 2018-09-27 | 2022-12-06 | Mitsubishi Electric Corporation | Left-behind detection device and left-behind detection method |
| CN109484292B (en) * | 2018-10-15 | 2021-11-30 | 上海理工大学 | Detection system and detection method for automatically detecting passengers detained in school bus |
| CN109558848A (en) * | 2018-11-30 | 2019-04-02 | 湖南华诺星空电子技术有限公司 | A kind of unmanned plane life detection method based on Multi-source Information Fusion |
| CN110827317B (en) * | 2019-11-04 | 2023-05-12 | 西安邮电大学 | Four-eye moving object detection and identification equipment and method based on FPGA |
| CN113361410B (en) * | 2021-06-07 | 2022-08-26 | 上海数川数据科技有限公司 | Child detection box filtering algorithm based on grid clustering |
| CN113997898B (en) * | 2021-11-30 | 2022-10-14 | 浙江极氪智能科技有限公司 | Living body detection method, apparatus, device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1698381A (en) * | 2003-05-08 | 2005-11-16 | 西门子公司 | Method and device for detecting an object or a person |
| CN101520850A (en) * | 2009-04-17 | 2009-09-02 | 中国科学院计算技术研究所 | Construction method of object detection classifier, object detection method and corresponding system |
| US8886206B2 (en) * | 2009-05-01 | 2014-11-11 | Digimarc Corporation | Methods and systems for content processing |
-
2016
- 2016-05-13 CN CN201610314230.4A patent/CN106022235B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1698381A (en) * | 2003-05-08 | 2005-11-16 | 西门子公司 | Method and device for detecting an object or a person |
| CN101520850A (en) * | 2009-04-17 | 2009-09-02 | 中国科学院计算技术研究所 | Construction method of object detection classifier, object detection method and corresponding system |
| US8886206B2 (en) * | 2009-05-01 | 2014-11-11 | Digimarc Corporation | Methods and systems for content processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106022235A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106022235B (en) | Missing child detection method based on human body detection | |
| TWI759286B (en) | System and method for training object classifier by machine learning | |
| CN109460699B (en) | Driver safety belt wearing identification method based on deep learning | |
| CN101655907B (en) | Trainman driving state monitoring intelligent alarm system | |
| TWI430212B (en) | Abnormal behavior detection system and method using automatic classification of multiple features | |
| CN109040693B (en) | Intelligent alarm system and method | |
| Wang et al. | Video smoke detection using shape, color and dynamic features | |
| CN104866842B (en) | A kind of human body target intrusion detection method | |
| CN102426757A (en) | Safety driving monitoring system based on mode identification and method thereof | |
| CN201307337Y (en) | Automatic alarming and monitoring device for traffic-lights nonobservance of pedestrian | |
| CN107578012A (en) | A kind of drive assist system based on clustering algorithm selection sensitizing range | |
| CN106096504A (en) | A kind of model recognizing method based on unmanned aerial vehicle onboard platform | |
| CN202771441U (en) | Face recognition monitoring system | |
| WO2022161139A1 (en) | Driving direction test method and apparatus, computer device, and storage medium | |
| WO2017098265A1 (en) | Method and apparatus for monitoring | |
| CN105809890A (en) | School-bus-safety-oriented missed-child detecting method | |
| CN104299363A (en) | Fatigue driving pre-warning system based on multi-feature fusion | |
| CN109657555A (en) | It is a kind of to get off the detection device and method of omission for school bus student | |
| CN111317653B (en) | Interactive intelligent auxiliary device and method for blind person | |
| Sefat et al. | Implementation of vision based intelligent home automation and security system | |
| CN115410324A (en) | Car as a house night security system and method based on artificial intelligence | |
| CN109800640A (en) | A kind of smoking detection method based on Two-dimensional Surfaces fitting | |
| CN109800686A (en) | A kind of driver's smoking detection method based on active infrared image | |
| CN105206060A (en) | Vehicle type recognition device and method based on SIFT characteristics | |
| CN111209814B (en) | Face recognition campus area early warning method and system for K12 education stage |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |