CN117238100A - Intelligent monitoring method and system for warehouse safety based on image recognition - Google Patents
Intelligent monitoring method and system for warehouse safety based on image recognition Download PDFInfo
- Publication number
- CN117238100A CN117238100A CN202311241183.1A CN202311241183A CN117238100A CN 117238100 A CN117238100 A CN 117238100A CN 202311241183 A CN202311241183 A CN 202311241183A CN 117238100 A CN117238100 A CN 117238100A
- Authority
- CN
- China
- Prior art keywords
- image
- gait
- field personnel
- personal computer
- industrial personal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000012544 monitoring process Methods 0.000 title claims abstract description 30
- 230000005021 gait Effects 0.000 claims abstract description 84
- 238000001514 detection method Methods 0.000 claims abstract description 33
- 206010017577 Gait disturbance Diseases 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 28
- 230000008859 change Effects 0.000 claims description 22
- 238000012706 support-vector machine Methods 0.000 claims description 18
- 230000007704 transition Effects 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 8
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 description 12
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011410 subtraction method Methods 0.000 description 2
- IAYPIBMASNFSPL-UHFFFAOYSA-N Ethylene oxide Chemical compound C1CO1 IAYPIBMASNFSPL-UHFFFAOYSA-N 0.000 description 1
- 208000034819 Mobility Limitation Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 230000037396 body weight Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005283 ground state Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application discloses a storage safety intelligent monitoring method and system based on image recognition. The industrial personal computer internally runs with an image processing unit, a fall detection unit and a gait recognition unit. The video acquisition module acquires a camera image sequence in real time and transmits the camera image sequence to the industrial personal computer, the image preprocessing unit in the industrial personal computer processes the image sequence and transmits the image sequence to the fall detection unit and the gait recognition unit respectively, the fall detection and the gait recognition judgment are carried out on the profile data of the field personnel respectively, and if the fall situation or abnormal gait is judged, a safety hazard signal is sent. The application has the advantages that whether the accident happens when the personnel falls down on site or not can be detected remotely, the suspicious personnel on site can be identified, the early warning and warning can be sent out in time, and the safety of the personnel on site and the property safety of the warehouse are improved.
Description
Technical Field
The application relates to the technical field of warehouse management, in particular to a warehouse safety intelligent monitoring method and system based on image identification.
Background
Logistics storage safety is a production operation link of an enterprise. Currently, in the security management of logistics warehouse, a manager is required to monitor video equipment in real time. After an abnormal situation occurs, the generation process of unsafe behavior of a target cannot be effectively early-warned in time in the past by playing back the monitoring video. And due to the physiological weakness of the person, inattention often occurs when monitoring the system. Traditional video security monitoring in a warehouse area cannot meet the increasing security requirements of enterprises, and development is being sought in the direction of machine learning-based video monitoring.
Because the storage environment is complex and the space is large, the monitoring camera is generally arranged at a higher position of the warehouse, the remote identification of the identity of the on-site staff is difficult by adopting the traditional target detection method, and the phenomena of extraneous staff, theft and the like are easy to occur; the existing warehouse safety management mainly relies on real-time monitoring of management staff on a video society, and because of a plurality of scenes, the phenomena of falling injury and physical state of staff are difficult to discover in time, and secondary disasters and the like are easy to cause; warehouse generally stores a large amount of materials and is easy to cause phenomena such as external personnel, theft and the like.
Disclosure of Invention
Based on the problems, the application provides a storage safety intelligent monitoring method and system based on image recognition, which can remotely detect whether a falling accident occurs to the field personnel or not and can identify the field suspicious personnel, detect whether the field personnel is injured or uncomfortable, send out early warning in time and improve the safety and property safety of the storage field personnel through detecting and identifying the storage field personnel. The application adopts the following technical scheme.
The intelligent monitoring method for the warehouse safety based on the image identification is characterized by comprising the following steps of:
s1: acquiring an image sequence of a video acquisition module in real time, and detecting targets of field personnel;
s2: performing data processing on the field personnel image sequence to obtain binarized field personnel profile data;
s3: safety risk identification is carried out on the field personnel based on the profile data, and if the safety risk exists, a safety risk signal is sent out;
s4: and sending out an early warning prompt according to the risk signal, and starting to store the current image.
Preferably, the data processing method for the field personnel image sequence in the step 2 is as follows:
background modeling is carried out by adopting a method of intermediate value, an original image sequence image function is subtracted from a reconstructed background image function, a threshold value is selected as a boundary for distinguishing the background from a moving target, if the threshold value is larger than the threshold value, the object is the background, original contour data of field personnel are obtained, and then a morphological operator is adopted for filtering noise and filling small holes on the original contour data.
Preferably, step S3 of safety risk identification includes performing fall detection on profile data of the field personnel,
the method comprises the following specific steps:
s31: judging whether the aspect ratio of the external rectangle of the human body is larger than 1, if so, entering the next step, otherwise, returning to the step S1;
s32: judging whether the ratio of the effective areas of the human body is larger than a threshold value, if so, entering the next step, otherwise, returning to the step S1;
s33: and judging whether the change rate K of the mass center of the human body is 0, if so, sending a safety risk signal, and otherwise, returning to the step S1.
Preferably, the step S3 of identifying the safety risk includes performing a fall detection on profile data of the field personnel, including:
acquiring the ratio of the width value (W) to the height value (H) of the minimum circumscribed rectangle of the human body outline of the field personnel through obtaining the outline data of the field personnel, namely the aspect ratio: WH (Wireless energy System) Ratio ＝W/H；
Calculating the ratio of the occupied area (Sp) of the human body to the area (SR) of the rectangle with the outermost contour in the rectangle with the outermost contour through the contour data of the field personnel, namely the effective area ratio EA of the human body Ratio ＝Sp/SR；
Acquiring human body centroid coordinates of each frame of on-site personnel outline, and calculating human body centroid change rate K= (y) of human body centroid coordinates A and B of two frames before and after calculation of adjacent data 1 -y 0 )/|(x 1 -x 0 )|，A＝(x 0 ,y 0 )，B＝(x 1 ,y 1 )。
Preferably, the step S3 of safety risk identification includes gait identification of field personnel, and the specific process is as follows:
and (3) extracting gait cycle feature vectors through field personnel profile data, inputting the feature vectors into a support vector machine for gait classification detection, and sending a safety risk signal when abnormal gait is detected, otherwise, returning to the step (1).
Preferably, the method for extracting the gait feature vector comprises the following steps:
drawing a gait cycle characteristic change curve by acquiring the ratio of the distance between two feet and the height in the side profile data of the field personnel in each frame of image, and selecting a characteristic vector G= [ C ] according to the characteristic curve y PH 1 PH 2 W h W L A h A L P f T 1 T 2 ]，C y For gait cycle, PH 1 For the first half cycle of gait, PH 2 Is the half period of gait, W h For peak width, W L Is the width of the trough A h For peak amplitude, A L Is the amplitude of the trough, P f For peak variance, T 1 For the transition time of the valley peak, T 2 Is the peak-to-valley transition time.
Preferably, the method for obtaining the trained support vector machine comprises the following steps:
and inputting the normal gait feature vector and the abnormal gait feature vector training data set into a support vector machine for training to obtain a trained support vector machine.
The intelligent monitoring system for warehouse safety based on image recognition comprises an image acquisition module, an industrial personal computer and an alarm module, wherein the upper computer operates a display module; an image processing unit, a fall detection unit and a gait recognition unit are arranged in the industrial personal computer; the image acquisition module and the upper computer operation display module are connected with the industrial personal computer through a hard wire; the image pre-processing unit is respectively connected with the fall detection unit and the gait recognition unit through software interfaces;
the image acquisition module acquires an image sequence in real time, performs target detection on the field personnel, transmits the image sequence to the industrial personal computer, and performs data processing on the image sequence by an image preprocessing unit in the industrial personal computer to obtain binarized field personnel profile data; the outline data of the field personnel are respectively transmitted to the fall detection unit and the gait recognition unit, fall detection and gait recognition judgment are respectively carried out on the outline data of the field personnel, if the fall situation or abnormal gait is judged, a safety risk signal is sent to the alarm module through a hard wire of the industrial personal computer, meanwhile, the industrial personal computer starts to store the current image, and the alarm module sends out an alarm sound when receiving the safety risk signal of the industrial personal computer; the upper computer operation display module displays the image acquired by the image acquisition module in real time, and can operate the industrial personal computer to read the stored image.
Compared with the prior art, the application has the following advantages:
target identification and danger judgment can be carried out on warehouse personnel in a long distance. Can in time early warning to unsafe conditions such as fall down of staff in time through falling down detection, refine the judgement process simultaneously, judge again in time when intermediate course does not satisfy, can save the power of calculation to correct recognition result, improved the degree of accuracy of discernment. On the other hand, based on the analysis of gait characteristic parameters, through gait recognition, the physical condition of on-site staff can be recognized and judged, and meanwhile, the theft behavior can be early warned.
Drawings
Fig. 1 is a schematic diagram of a module of a warehouse safety intelligent monitoring system based on image recognition.
Fig. 2 is a flow chart of a method for intelligent monitoring of warehouse safety based on image recognition.
Fig. 3 is a flowchart of a fall detection method.
Fig. 4 is a flow chart of a gait recognition process based on a support vector machine.
FIG. 5 is a gait characteristic change curve for a normal gait cycle
FIG. 6 is a gait characteristic change curve for an abnormal gait cycle
FIG. 7 is a schematic view of characteristic parameters of a gait cycle gait characteristic change curve
Detailed Description
The following detailed description is exemplary and is intended to provide further explanation of the 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. It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application.
Example 1
The warehouse is monitored in real time by using management personnel, and the phenomena of secondary disasters and the like are difficult to discover in time due to more scenes and the fact that the personnel fall down and are injured and the physical state is poor; because the storage environment is complex, the space is larger, the monitoring camera is generally arranged at a higher position of the warehouse, and the remote identification of the identity of the on-site staff is difficult. The application provides a storage safety intelligent monitoring system based on image recognition, which is shown in fig. 1, and can be used for recognizing the falling danger situation of field personnel. An image processing unit and a fall detection unit are operated in the industrial personal computer; the video acquisition module camera and the upper computer operation display module are connected with the industrial personal computer through wires. The image preprocessing unit is connected with the gait recognition unit through an internal software interface of the industrial personal computer.
A flow chart of a method for intelligent monitoring of warehouse safety based on image recognition is shown in fig. 2, specifically:
s1, an image acquisition module acquires an image sequence in real time, performs target detection on field personnel, and transmits the image sequence to an industrial personal computer;
s2, an internal image preprocessing unit of the industrial personal computer performs data processing on the image sequence to obtain binarized field personnel profile data; transmitting the profile data of the field personnel to a fall detection unit;
the specific image preprocessing unit adopts a background subtraction method to carry out background modeling through a method of an intermediate value, an original image sequence image function is subtracted from a reconstructed background image function, a threshold value is selected as a boundary for distinguishing the background from a moving object, if the threshold value is larger than the threshold value, the object is the background, and if the threshold value is smaller than the boundary, the original scene personnel contour data is obtained, and then a morphological operator is adopted to filter noise from the original contour data and fill a small hole.
The specific method is as follows, background modeling is carried out by adopting an RGB space background modeling method, intermediate values of three components R, G and B of n frames of images are taken as pixel values of the background images, and the background images are as follows:
B(x,y)＝med{f(k,x,y).c}
f (k, x, y), k=1, 2, … … n represents a sequence of acquired gait color image functions, B (x, y) represents RGB vectors at the established background point (x, y), and c represents one of the RGB color components.
After the background model is obtained, the difference function f (a, b) is obtained by subtracting the current image from the background image, and the difference function is as follows according to experience:
0≤f(a,b)＜1,0≤a(x,y),b(x,y)≤255
where a (x, y) and b (x, y) are luminance values of the current image and the background image at the pixel (x, y), respectively.
And obtaining an original binarized image by selecting a threshold value T and adopting the following formula.
When the color of the human body area and the color of the background are not greatly different, the threshold value can be set smaller, and conversely, the threshold value needs to be set larger. Due to interference and noise, the binarized human body contour data is subjected to morphological operator to further filter the noise and fill small holes, so that the final binarized field human body contour data is obtained.
S3, carrying out safety risk identification on the field personnel based on the contour data by the fall detection unit, specifically carrying out fall detection on the contour data of the field personnel, and sending a safety risk signal to the alarm module through the industrial personal computer hard line if the fall situation is judged to exist;
the specific process of fall detection shown in fig. 3 is:
s11: judging whether the aspect ratio of the external rectangle of the human body is larger than 1, if so, entering the next step, otherwise, returning to the step S1;
s12: judging whether the ratio of the effective areas of the human body is larger than a threshold value, if so, entering the next step, otherwise, returning to the step S1;
s13: and judging whether the human centroid change rate K is 0, if so, sending a security risk alarm signal, otherwise, returning to the step S1.
The reason why the above-described judgment conditions are selected in the fall detection process in this embodiment is given below for explanation, and the definition thereof is set forth:
because the minimum circumscribed rectangle of human body contour, the ratio of width to height is less than 1 when the people stand normally, when the people fall down, the ratio of width to height of minimum circumscribed rectangle of human body contour is greater than 1. Therefore, the ratio of the width to the height of the rectangle circumscribed by the human body outline can reflect whether the person falls down or not.
Therefore, the width (W) and the height (H) of the rectangle circumscribed by the human body are obtained according to the obtained human body outline data.
However, judgment by only this feature causes erroneous judgment, and thus it is necessary to add judgment conditions to correct it next.
(1) For example, the ratio of the width to the height of the field staff member when bending down to carry the goods may be larger than 1, but because the size of the human body is determined, when the field staff member bends down to work, the external rectangular area of the human body becomes larger, so that the ratio of the area (Sp) occupied by the human body to the outermost rectangular area (SR) becomes smaller, and when the ratio is smaller than a specified threshold value, the field staff member can be considered not to fall down.
It is therefore also necessary to calculate the ratio of the area occupied by the human body (Sp) to the area of the outermost contour rectangle (SR), namely: EA (EA) Ratio =sp/SR. In EA (ethylene oxide) Ratio Sp represents the number of pixels with 1 pixel value in the rectangular frame, namely the number of pixels in the human body outline, and SR represents the number of all pixels in the whole rectangular frame.
(2) In addition, if the worker stands up after falling down, the worker is considered not to need early warning at the moment, so that the change of the mass center of the human body is further corrected by judging. According to the obtained personAnd calculating the change rate of the mass center of the human body of the front and rear frame images to obtain the body contour data. Let A and B be the centroid coordinates of two adjacent frames, respectively, which are (x) 0 ,y 1 ) And (x) 1 ,y 1 ) The centroid rate of change is then:
K＝(y 1 -y 0 )/|(x 1 -x 0 )|
in the formula, K is the centroid change rate, and when K is a positive value, the human body moves upwards, namely the measured object moves upwards, and the measured object stands up even in the inverted ground state. Indicating that the tumbling person is not at risk, which is not required to be alerted. When K is 0, the measured object does not move in the vertical direction, if the measured object is in a falling state, the measured object indicates that a falling person cannot stand up in a short time and needs the help of others, and the monitoring system needs to give an alarm. When K is negative, it indicates that the detected object is falling, and it is likely to fall or make some movements, and the next frame needs to be determined.
And S4, the alarm module sends out an early warning prompt when receiving a safety risk signal of the industrial personal computer, and the industrial personal computer starts to store the video in the memory so that a worker can conveniently call the stored image through the upper computer operation display module.
The upper computer operation display module can display the image acquired by the image acquisition module in real time.
Example two
The warehouse is monitored in real time by using management personnel, and the phenomena of secondary disasters and the like are difficult to discover in time due to more scenes and the fact that the personnel fall down and are injured and the physical state is poor; because the storage environment is complex, the space is larger, the monitoring camera is generally arranged at a higher position of the warehouse, and the remote identification of the identity of the on-site staff is difficult. Therefore, the intelligent monitoring system for warehouse safety based on image recognition, as shown in fig. 1, can be used for early warning on abnormal physical conditions of field personnel through gait recognition. The system comprises a video acquisition module, an industrial personal computer, an alarm module and an upper computer operation display module. The industrial personal computer is internally provided with an image processing unit and a gait recognition unit. The video acquisition module camera and the upper computer operation display module are connected with the industrial personal computer through wires. The image preprocessing unit is connected with the gait recognition unit through an internal software interface of the industrial personal computer.
A flow chart of a method for intelligent monitoring of warehouse safety based on image recognition is shown in fig. 2, and the specific process is as follows:
s1, an image acquisition module acquires an image sequence in real time, performs target detection on field personnel, and transmits the image sequence to an industrial personal computer;
s2, an internal image preprocessing unit of the industrial personal computer performs data processing on the image sequence, and detects and identifies field personnel to obtain binarized field personnel profile data; transmitting the profile data of the field personnel to a gait recognition unit;
the image preprocessing unit adopts a background subtraction method to carry out background modeling through a method of an intermediate value, subtracts an original image sequence image function from a reconstructed background image function, selects a threshold value as a boundary for distinguishing a background from a moving object, takes the object as the object when the threshold value is larger than the threshold value, and takes the object as the background when the threshold value is smaller than the boundary, so that original contour data of field personnel are obtained. And then adopting a morphological operator to filter noise and fill small holes from the original contour data.
The specific method comprises the following steps: performing background modeling by adopting an RGB space background modeling method, taking the intermediate values of three components R, G and B of n frames of images as pixel values of the background image, and then the background image is:
B(x,y)＝med{f(k x y).c}
f (k, x, y), k=1, 2, … … n represents a sequence of acquired gait color image functions, B (x, y) represents RGB vectors at the established background point (x, y), and c represents one of the RGB color components.
After the background model is obtained, the difference function f (a, b) is obtained by subtracting the current image from the background image, and the difference function is as follows according to experience:
0≤f(a,b)＜1,0≤a,b≤255
where a and b are the luminance values of the current image and the background image at the pixel (x, y), respectively.
And obtaining an original binarized image by selecting a threshold value T and adopting the following formula.
When the color of the human body area and the color of the background are not greatly different, the threshold value can be set smaller, and conversely, the threshold value needs to be set larger. Due to interference and noise, the binarized human body contour data is subjected to morphological operator to further filter the noise and fill small holes, so that the final binarized field personnel human body contour image data is obtained.
S3, safety risk recognition is carried out on the field personnel based on the profile data by the gait recognition unit, specifically, the gait cycle characteristic vector is extracted through the profile data of the field personnel, the characteristic vector is input into a support vector machine for gait classification detection, and when abnormal gait is detected, a safety risk signal is sent out, otherwise, the step 1 is returned.
The method adopted by the S31 gait recognition unit is a method (Support Vector Machine, SVM) based on the prior art support vector machine, which can carry out single-classification detection, namely, the trained support vector machine is obtained by inputting a normal gait feature vector and an abnormal gait feature vector data set into the support vector machine for training, and the method can carry out classification detection on the input gait feature vector and judge whether the gait feature vector is abnormal or not.
The method for extracting the gait feature vector in S32 comprises the following steps: drawing a gait cycle characteristic change curve by acquiring the ratio of the distance between two feet and the height in the side profile data of the field personnel in each frame of image, and selecting a characteristic vector G= [ C ] according to the characteristic curve y PH 1 PH 2 W h W L A h A L P f T 1 T 2 ]，C y For gait cycle, PH 1 For the first half cycle of gait, PH 2 Is the half period of gait, W h For peak width, W L Is the width of the trough A h Is the peak amplitude, A is the trough amplitude, P f For peak variance, T 1 For the transition time of the valley peak, T 2 Is the peak-to-valley transition time.
The reason for extracting the gait feature vector is analyzed as follows: firstly, the gait motion process of a person needs to be analyzed in the process of extracting the gait feature vector, wherein the gait motion of the person is a periodic motion, and one gait cycle refers to a period of time when the same foot is in heel strike twice in succession. The different feet are grounded in one step at successive heel angles, and one gait cycle comprises two steps. The change with time of the distance between the two feet when the person walks can be extracted as a parameter describing gait while taking into account the influence of the image size. Therefore, the feature quantity is defined as the ratio of the distance(s) between two feet of the profile data of the field personnel in each frame of image to the height (h): r=s/h as a parameter describing gait. The gait characteristic change curve of a normal gait cycle is shown as fig. 5, the gait characteristic change curve can be shown as fig. 6 when abnormality occurs, the abscissa in the figure is the frame number, the ordinate is the R, the peak value of the ratio R of two adjacent steps in the abnormal gait characteristic change curve is obviously different greatly, the difference value is more than 0.05, and the situation that the pedestrian possibly gets injured is reflected; the ratio R of the two steps before and after the normal gait cycle does not change much. Therefore, the gait characteristic change curve of the gait parameters can be used for distinguishing the normal gait from the abnormal gait
Further, in order to better describe gait characteristics, several characteristic parameters are defined and how the characteristic parameters characterize abnormal gait is specifically analyzed. As shown in the graph of FIG. 7, the abscissa is the frame number, and the ordinate is the ratio R, which is a gait cycle with two adjacent peaks.
Gait cycle (C) y ) Defined as the time between the ith peak and the (i+2) th peak, which characterizes the walker's walking speed, if the gait cycle is longer, the on-site person is considered to walk slowly, and can be considered to be physically under-riddenGood or tired of working. If the gait cycle is fast, it is indicated that the pedestrian is urgent, and an emergency or theft may be suspected.
Gait half-cycle (PH) 1 ) And the second half Period (PH) 2 ). Refers to the time to support body weight with either the right foot or the left foot during a gait cycle, defined as the time between the i-th peak to the (i+1) -th peak and the (i+1) -th peak to the (i+2) -th peak, respectively. Normally both are about equal. If the two are not equal, it is indicated that a certain leg or foot of the field personnel is problematic, i.e. the field personnel is injured or physically uncomfortable.
Peak width (W) h ) And trough width (W) L ). The wave crest width refers to the time of two feet simultaneously touching the ground in the walking process of the field personnel, and represents the conversion capability of the field personnel from the balanced state of the two feet simultaneously touching the ground to the unbalanced state of the single foot. If the peak is too wide, it is an indication that the field personnel is not careful when making the transition from the equilibrium to the non-equilibrium state. The width of the trough represents the short time of the two legs being folded in the walking process, if the trough is too wide, the on-site personnel must land and rest slightly when the two legs are folded so as to solve the problem of insufficient supporting capacity of the supporting foot.
Peak amplitude A h And trough amplitude A L The peak amplitude refers to the peak of the curve, which represents the step size of a person, and if the value is too small in the curve, it means that the legs and feet of a person on site may be problematic to step up. The valley amplitude refers to the value of the valley position, which represents the width of the overlapping legs, and if this value is too great, it indicates that the field personnel may have difficulty walking.
Peak variance (P) f ) The change condition of the stride can be reflected, and the problem of the legs and feet of the on-site personnel is reflected when the stride is uneven. P (P) f The calculation formula of (2) is as follows:wherein->For a period ofThe average of all peaks in the interval.
Valley-peak transition time (T) 1 ) And peak-to-valley transition time (T) 2 ) The valley-peak transition time refers to the time from overlapping of two legs to simultaneous touchdown of two feet of a walker, and the single foot movement speed of a field worker in the time can be described by combining the peak amplitude. When the peak amplitude is larger and the valley value is smaller, the movement speed of the single foot is high, even running, emergency or unusual occurrence is proved, otherwise, when the amplitude is larger and the transition time is longer, the pedestrian is proved to walk slowly, and the sounding or attention-drawing suspicion is avoided. Similarly, the peak-to-valley transition time is used to describe the speed of movement of a single foot during the time from when both feet are simultaneously touching the ground to when both feet overlap.
Therefore, the characteristic parameters can better reflect the normal and dangerous conditions of field personnel in the warehouse, and the characteristic vector G= [ C ] is obtained according to the data extracted from each frame of image in one gait cycle y PH 1 PH 2 W h W L A h A L P f T 1 T 2 ]。
As shown in fig. 4, the support vector machine based gait recognition process includes two steps. The first part is a gait training part, firstly, an image sequence acquired from a video is processed, normal gait and abnormal gait feature vectors are respectively collected to be used as gait recognition data, then the data are used as a training sample set for gait recognition, and the data are input into a support vector machine for training, so that a trained support vector machine model is obtained. And in the second part, the identification part also reads in the image sequence acquired by the video acquisition module, processes the image sequence to obtain a feature vector, inputs the feature vector into a trained support vector machine, and finally outputs whether the result is abnormal gait.
And S4, the alarm module sends out an early warning prompt when receiving a safety risk signal of the industrial personal computer, and the industrial personal computer starts to store the video in the memory so that a worker can conveniently call the stored image through the upper computer operation display module.
The upper computer operation display module can display the image acquired by the image acquisition module in real time.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The foregoing is merely a preferred embodiment of the present application, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present application, and such modifications and variations should also be regarded as being within the scope of the application.
Claims (8)
1. The intelligent monitoring method for the warehouse safety based on the image identification is characterized by comprising the following steps of:
s1: acquiring an image sequence of a video acquisition module in real time, and detecting targets of field personnel;
s2: performing data processing on the field personnel image sequence to obtain binarized field personnel profile data;
s3: safety risk identification is carried out on the field personnel based on the profile data, and if the safety risk exists, a safety risk signal is sent out;
s4: and sending out an early warning prompt according to the risk signal, and starting to store the current image.
2. The method for intelligently monitoring warehouse safety based on image recognition as claimed in claim 1, wherein the step 2 of performing data processing on the field personnel image sequence is as follows:
background modeling is carried out by adopting a method of intermediate value, an original image sequence image function is subtracted from a reconstructed background image function, a threshold value is selected as a boundary for distinguishing the background from a moving target, if the threshold value is larger than the threshold value, the object is the background, original contour data of field personnel are obtained, and then a morphological operator is adopted for filtering noise and filling small holes on the original contour data.
3. The intelligent monitoring method for warehouse safety based on image recognition according to claim 1, wherein the step S3 of safety risk recognition comprises the steps of performing fall detection on profile data of field personnel, and the specific steps are as follows:
s31: judging whether the aspect ratio of the external rectangle of the human body is larger than 1, if so, entering the next step, otherwise, returning to the step S1;
s32: judging whether the ratio of the effective areas of the human body is larger than a threshold value, if so, entering the next step, otherwise, returning to the step S1;
s33: and judging whether the change rate K of the mass center of the human body is 0, if so, sending a safety risk signal, and otherwise, returning to the step S1.
4. The method for intelligent monitoring of warehouse safety based on image recognition as set forth in claim 3, wherein the step S3 of safety risk recognition includes a fall detection of profile data of a person on site, comprising:
acquiring the ratio of the width (W) to the height (H), i.e. the aspect ratio WH, of the minimum bounding rectangle of the human body profile of the site person from the site person profile data Ratio ＝W/H；
Calculating the ratio of the occupied area (Sp) of the human body to the area (SR) of the rectangle with the outermost contour in the rectangle with the outermost contour through the contour data of the field personnel, namely the effective area ratio EA of the human body Ratio ＝Sp/SR；
Acquiring human body centroid coordinates of each frame of on-site personnel outline, and calculating human body centroid change rate K= (t) of human body centroid coordinates A and B of two frames before and after calculation of adjacent data 1 -y 0 )/|(x 1 -x 0 )|，A＝(x 0 ,y 0 )，B＝(x 1 ,y 1 )。
5. The method for intelligently monitoring warehouse safety based on image recognition according to claim 1, wherein the step S3 of safety risk recognition comprises gait recognition of field personnel, and the specific process is as follows:
and (3) extracting gait cycle feature vectors through field personnel profile data, inputting the feature vectors into a trained support vector machine for gait classification detection, and sending a safety risk signal when abnormal gait is detected, otherwise, returning to the step (1).
6. The method for intelligently monitoring warehouse safety based on image recognition according to claim 5, wherein the method for extracting gait feature vectors is as follows:
acquiring the ratio of the distance between two feet and the height in the side profile data of the field personnel in each frame of image, drawing a gait cycle characteristic change curve, and selecting a characteristic vector G= [ C ] according to the characteristic curve y PH 1 PH 2 W h W L A h A L P f T 1 T 2 ]，C y For gait cycle, PH 1 For the first half cycle of gait, PH 2 Is the half period of gait, W h For peak width, W L Is the width of the trough A h For peak amplitude, A L Is the amplitude of the trough, P f For peak variance, T 1 For the transition time of the valley peak, T 2 Is the peak-to-valley transition time.
7. The method for intelligently monitoring warehouse safety based on image recognition as set forth in claim 6, wherein the method for obtaining the trained support vector machine is as follows:
and inputting the normal gait feature vector and the abnormal gait feature vector training data set into a support vector machine for training to obtain a trained support vector machine.
8. The intelligent monitoring system for warehouse safety based on image recognition comprises an image acquisition module, an industrial personal computer and an alarm module, wherein the upper computer operates a display module; an image processing unit, a fall detection unit and a gait recognition unit are arranged in the industrial personal computer; the image acquisition module and the upper computer operation display module are connected with the industrial personal computer through a hard wire; the image pre-processing unit is respectively connected with the fall detection unit and the gait recognition unit through software interfaces;
the image acquisition module acquires an image sequence in real time, performs target detection on the field personnel, transmits the image sequence to the industrial personal computer, and performs data processing on the image sequence by an image preprocessing unit in the industrial personal computer to obtain binarized field personnel profile data; the outline data of the field personnel are respectively transmitted to the fall detection unit and the gait recognition unit, fall detection and gait recognition judgment are respectively carried out on the outline data of the field personnel, if the fall situation or abnormal gait is judged, a safety risk signal is sent to the alarm module through a hard wire of the industrial personal computer, meanwhile, the industrial personal computer starts to store the current image, and the alarm module sends out an alarm sound when receiving the safety risk signal of the industrial personal computer; the upper computer operation display module displays the image acquired by the image acquisition module in real time, and can operate the industrial personal computer to read the stored image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311241183.1A CN117238100A (en) | 2023-09-25 | 2023-09-25 | Intelligent monitoring method and system for warehouse safety based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311241183.1A CN117238100A (en) | 2023-09-25 | 2023-09-25 | Intelligent monitoring method and system for warehouse safety based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117238100A true CN117238100A (en) | 2023-12-15 |
Family
ID=89098082
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311241183.1A Pending CN117238100A (en) | 2023-09-25 | 2023-09-25 | Intelligent monitoring method and system for warehouse safety based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117238100A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117953639A (en) * | 2023-12-25 | 2024-04-30 | 上海驰瑞云智能科技有限公司 | Intelligent home security system and method based on Internet of things |
-
2023
- 2023-09-25 CN CN202311241183.1A patent/CN117238100A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117953639A (en) * | 2023-12-25 | 2024-04-30 | 上海驰瑞云智能科技有限公司 | Intelligent home security system and method based on Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10810414B2 (en) | Movement monitoring system | |
| JP6534499B1 (en) | MONITORING DEVICE, MONITORING SYSTEM, AND MONITORING METHOD | |
| CN103576660B (en) | Smart Home method for supervising | |
| US10482613B2 (en) | Movement monitoring system | |
| CN112165600B (en) | Drowning identification method and device, camera and computer system | |
| CN103942850A (en) | Medical staff on-duty monitoring method based on video analysis and RFID (radio frequency identification) technology | |
| CN117238100A (en) | Intelligent monitoring method and system for warehouse safety based on image recognition | |
| CN109145696B (en) | Old people falling detection method and system based on deep learning | |
| JP5598751B2 (en) | Motion recognition device | |
| KR101708233B1 (en) | Apparaus and method for detecting aggressive actions from streaming 3d skeletton data | |
| CN111242004A (en) | Automatic alarm method and system based on elevator monitoring data processing | |
| CN112163477A (en) | Escalator pedestrian pose target detection method and system based on FasterR-CNN | |
| EP2000998B1 (en) | Flame detecting method and device | |
| CN114842560B (en) | Computer vision-based construction site personnel dangerous behavior identification method | |
| CN113506416A (en) | Engineering abnormity early warning method and system based on intelligent visual analysis | |
| KR101454644B1 (en) | Loitering Detection Using a Pedestrian Tracker | |
| TWI493510B (en) | Falling down detection method | |
| CN112798979A (en) | System and method for detecting state of grounding wire of transformer substation based on deep learning technology | |
| CN114299106A (en) | High-altitude parabolic early warning system and method based on visual sensing and track prediction | |
| CN114155492A (en) | High-altitude operation safety belt hanging rope high-hanging low-hanging use identification method and device and electronic equipment | |
| CN113449675A (en) | Coal mine personnel border crossing detection method | |
| CN114359825A (en) | Monitoring method and related product | |
| CN111597889B (en) | Method, device and system for detecting target movement in video | |
| CN115731563A (en) | Method for identifying falling of remote monitoring personnel | |
| KR102544492B1 (en) | Apparatus and method of managing safety of swimming pool |
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 |