CN112699769A - Detection method and system for left-over articles in security monitoring - Google Patents
Detection method and system for left-over articles in security monitoring Download PDFInfo
- Publication number
- CN112699769A CN112699769A CN202011564655.3A CN202011564655A CN112699769A CN 112699769 A CN112699769 A CN 112699769A CN 202011564655 A CN202011564655 A CN 202011564655A CN 112699769 A CN112699769 A CN 112699769A
- Authority
- CN
- China
- Prior art keywords
- key frame
- frame image
- network model
- image data
- monitoring video
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 80
- 238000012544 monitoring process Methods 0.000 title claims abstract description 74
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 230000003068 static effect Effects 0.000 claims abstract description 15
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 9
- 230000008859 change Effects 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Geometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of image processing, and discloses a method and a system for detecting a left article in security monitoring, wherein the method comprises the steps of preprocessing a monitoring video frame image data set; establishing a convolutional neural network model added with an SPP module; inputting a plurality of key frame images into a target detection network model, and outputting coordinate bounding boxes and labels of all objects in the key frame images by the target detection network model; and matching the same object in the front and back different key frame images by adopting a front and back frame minimum centroid distance matching method according to the centroid coordinates, judging whether the static time of the same object after being left over exceeds a time threshold, and outputting object left-over early warning information. The method has robustness and stability after the object is left for a long time; the invention not only improves the recall rate of small articles, but also quickly matches the same object based on the minimum distance of the mass center, has high matching accuracy and ensures that the articles can send out early warning information in time after being left.
Description
Technical Field
The invention relates to the field of image processing, in particular to a method and a system for detecting a left article in security monitoring.
Background
With the popularization and the wide use of network monitoring cameras, the detection of the left-over object is widely applied to the field of safety precaution, for example, in places such as banks, subways, railway stations, markets and the like, the left-over object is monitored and alarmed, and the practicability and the effectiveness of a video monitoring system are improved to a great extent. Although many researches are carried out on the detection of the remnants, some unsolved problems still exist, and a remnants detection algorithm suitable for various monitoring scenes does not exist, and meanwhile, an improvement space exists in the accuracy and the real-time performance of the detection of the remnants in a complex environment, so that an effective subway monitoring mechanism is helped to be established. At present, the detection of the remnant mainly comprises the following methods: a. a method for detecting an object based on a conventional image; b. an object detection method based on deep learning.
Based on the traditional image object detection method, most adopted methods are based on a Gaussian mixture model, for example, the national patent publication CN111914670A, which discloses a method, a device, a system and a storage medium for detecting a left article, the invention comprises the following steps: establishing a background model; acquiring a motion area in a detection image, and extracting a first characteristic point in the motion area; and acquiring a previous frame image of the detection image, and extracting a second characteristic point in an area which is overlapped with the motion area in the previous frame image. According to the method, the foreground object in the image is extracted by modeling the background of the video image, the static foreground object is timed, and if the timer reaches a threshold value, the foreground object is marked as a left article and an alarm is given. However, the detection using the background modeling method has the following two problems: firstly, the slower moving foreground object can be processed as background by mistake; secondly, stationary foreground objects are gradually absorbed by the background, and missing detection may be caused for stationary objects for a long time.
The invention discloses a detection method based on deep learning, such as national patent publication CN107527009B, and discloses a remnant detection method based on YOLO target detection, which detects a target in real time through a YOLO algorithm to obtain a target category and a corresponding specific coordinate in each frame of image data, classifies the detected target through the target category and the overlapping degree of the two coordinates, and judges a suspicious target by carrying out background removal after tracking and timing to obtain an accurate remnant. Although the invention can achieve the real-time effect, the detection effect on small target objects (such as bags) is poor.
The existing detection method for the left-over articles has the following main problems: in the traditional image detection object technology, the recall rate of foreground objects is low, and objects left for a long time are missed for detection. The existing detection technology based on the deep neural network has low recall rate on small objects in a scene and has poor detection effect on bags, suitcases and the like with small targets.
Disclosure of Invention
The invention provides a method and a system for detecting a left article in security monitoring, so as to solve the problems in the prior art.
In a first aspect, the invention provides a method and a system for detecting a left article in security monitoring, comprising the following steps:
s1) acquiring monitoring video frame image datasets in different scenes, preprocessing the monitoring video frame image datasets, setting a label for each image in the monitoring video frame image datasets, and acquiring a preprocessed monitoring video frame image dataset and a label dataset corresponding to the preprocessed monitoring video frame image dataset;
s2), establishing a target detection network model, wherein the target detection network model is a convolutional neural network model added with a Spatial Pyramid Pooling (SPP) module, and training the target detection network model according to a preprocessed monitoring video frame image data set and a label data set to obtain a trained target detection network model;
s3) collecting video stream image data of the to-be-detected legacy article, acquiring a plurality of key frame images in the video stream image data according to a time sequence, respectively inputting the plurality of key frame images into the trained target detection network model according to the time sequence, and respectively outputting coordinate bounding boxes and labels of all objects in each key frame image by the trained target detection network model;
s4) calculating the coordinates of the mass centers of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in the front and back different key frame images according to the coordinates of the mass centers and by adopting a front and back frame minimum mass center distance matching method, acquiring the remaining static time length of the same object in the matched front and back different key frame images, setting a time length threshold value, judging whether the remaining static time length of the same object in the matched front and back different key frame images exceeds the time length threshold value, and if not, returning to the step S3); if yes, outputting object leaving early warning information.
Further, in step S1), preprocessing the monitoring video frame image data set and setting a label for each image in the monitoring video frame image data set, including performing an enhancement preprocessing on each image data in the monitoring video frame image data set, where the enhancement preprocessing includes image cropping, image flipping and Mosaic data enhancement; the tags include pedestrians, bags, cartons and luggage.
Further, in step S2), the convolutional neural network model added with the SPP module includes an input layer, an intermediate layer, and an output layer; the middle layer comprises a plurality of convolution layers, a plurality of batch normalization layers, a plurality of nonlinear activation function layers and an SPP module; the output layer outputs object coordinate bounding boxes, confidence levels and labels in each image.
Further, in step S3), a plurality of key frame images in the video stream image data are acquired in time sequence, including setting a preset number of frames, and key frame images are acquired in the video stream image data every other preset number of frames in time sequence.
Further, in step S4), the same object in the previous and subsequent key frame images is matched according to the centroid coordinates and by using the previous and subsequent frame minimum centroid distance matching method, including performing matching association between different key frame images by using the minimum distance between the centroid coordinates of the object as a feature, and obtaining the same object located in different key frame images through the matching association.
Further, in step S4), obtaining the still time lengths of the same object left in the previous and subsequent different keyframe images, and setting a centroid coordinate change threshold value, calculating
Obtaining the jth same object in the front and back different key frame imagesCentroid coordinate difference ω in imagesjJ is less than or equal to m, m is the total number of the same object matched in the front and back different key frame images, and the centroid coordinate of the jth same object in the (i + 1) th key frame image is (x)i+1，yi+1) The centroid coordinate of the jth identical object in the ith key frame image is (x)i，yi) When the jth same object is in the centroid coordinate difference value omega in the front and back different key frame imagesjAnd when the change time is less than the centroid coordinate change threshold, judging the jth same object as the left-behind object, and acquiring the stationary time length of the jth same object after being left behind.
In a second aspect, the invention provides a system for detecting a left article in security monitoring, which comprises a monitoring video frame image acquisition module, a target detection network model establishing module and a left article detection module;
the monitoring video frame image acquisition module is used for acquiring monitoring video frame image datasets in different scenes and acquiring video stream image data of a to-be-detected article to be left, preprocessing the monitoring video frame image datasets, setting a label category for each image in the monitoring video frame image datasets, and acquiring a preprocessed monitoring video frame image dataset and a label dataset corresponding to the preprocessed monitoring video frame image dataset; acquiring a plurality of key frame images in video stream image data according to a time sequence;
the target detection network model establishing module is used for establishing a target detection network model, the target detection network model is a convolutional neural network model added into the SPP module, and the target detection network model is trained according to a monitoring video frame image data set and a label data set which are preprocessed in the monitoring video frame image acquisition module to obtain a trained target detection network model;
the legacy article detection module is used for respectively inputting a plurality of key frame images acquired by the monitoring video frame image acquisition module into the trained target detection network model according to the time sequence, and the trained target detection network model respectively outputs coordinate bounding boxes and labels of all objects in each key frame image; calculating the coordinates of the mass centers of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in front and back different key frame images according to the coordinates of the mass centers and by adopting a front and back frame minimum mass center distance matching method, acquiring the static time of the same object in the matched front and back different key frame images after the object is left, and outputting object left-behind early warning information according to the static time.
The invention has the beneficial effects that:
compared with the traditional image object detection method, the invention provides the object leaving method based on deep learning, the object and the pedestrian are detected by utilizing the computer vision technology, the detector which is stable for a long time is provided, and compared with the traditional background modeling method, the method has robustness and stability under the condition that the object is absorbed by the background after being left for a long time.
Compared with a detection method based on Darknet53, the method disclosed by the invention has the advantages that the recall rate of small articles is improved by performing data enhancement on the image and integrating the SPP module in front of the prediction characteristic diagram.
For the event of article leaving, the invention matches each object of different frames based on the minimum distance of the mass center, the method matches the same object quickly in the boundary frame information output by the detector, the matching accuracy is high, and the system can send out early warning information in time after the article is left.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments are briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for detecting a left-behind object in security monitoring according to a first embodiment of the present disclosure.
Fig. 2 is a schematic diagram of object matching relationship between different key frame images according to the first embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
An embodiment of a method and a system for detecting a left article in security monitoring is shown in fig. 1, and includes the following steps:
s1) acquiring surveillance video frame image datasets in different scenes, preprocessing the surveillance video frame image datasets, setting labels for each image in the surveillance video frame image datasets, and obtaining the preprocessed surveillance video frame image datasets and label datasets corresponding to the preprocessed surveillance video frame image datasets. Preprocessing a monitoring video frame image data set, setting a label for each image in the monitoring video frame image data set, and performing enhancement preprocessing on each image data in the monitoring video frame image data set, wherein the enhancement preprocessing comprises image cutting, image turning and a Mosaic data enhancement mode; the tags include pedestrians, bags, cartons and luggage.
S2), establishing a target detection network model, wherein the target detection network model is a convolutional neural network model added with an SPP module, and training the target detection network model according to the preprocessed monitoring video frame image data set and the label data set to obtain a trained target detection network model; the convolution neural network model added with the SPP module comprises an input layer, a middle layer and an output layer; the middle layer comprises a plurality of convolution layers, a plurality of batch normalization layers, a plurality of nonlinear activation function layers and an SPP module; the output layer outputs object coordinate bounding boxes, confidence levels and labels in each image.
S3) collecting video stream image data of the to-be-detected legacy article, acquiring a plurality of key frame images in the video stream image data according to a time sequence, respectively inputting the plurality of key frame images into the trained target detection network model according to the time sequence, and respectively outputting coordinate bounding boxes and labels of all objects in each key frame image by the trained target detection network model.
In step S3), a number of key frame images in the video stream image data are acquired in time series, including setting the preset number of frames to 4, and key frame images are acquired every 4 frames in the video stream image data in time series.
S4) calculating the coordinates of the mass centers of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in the front and back different key frame images according to the coordinates of the mass centers and by adopting a front and back frame minimum mass center distance matching method, acquiring the remaining static time length of the same object in the matched front and back different key frame images, setting a time length threshold value, judging whether the remaining static time length of the same object in the matched front and back different key frame images exceeds the time length threshold value, and if not, returning to the step S3); if yes, outputting object leaving early warning information.
In step S4), the same object in the previous and subsequent key frame images is matched according to the centroid coordinates and by using the matching method of the minimum centroid distance of the previous and subsequent frames, including performing matching association between different key frame images by using the minimum distance between the centroid coordinates of the object as a feature, and obtaining the same object located in different key frame images through the matching association. As shown in FIG. 2, the key frame image L of the previous framei-1The two detected articles are respectively article Ai-1And article Bi-1Current frame key frame image LiDetected two articles are article AiAnd article BiCalculating the coordinates of the mass center to obtain an article Bi-1And article BiHas a centroid distance of 0.1, and is an article Bi-1With article AiHas a centroid distance of 0.7, article Ai-1With article AiHas a centroid distance of 0.8, article Ai-1And article BiThe centroid distance between them is 0.3. In this example, article Bi-1And article BiThe centroid distance between them is minimized, and thus, item B is placedi-1And article BiConsidered as the same object in the front and back different key frame images。
In step S4), the method further includes obtaining the remaining static time lengths of the same object in the previous and subsequent different keyframe images, setting a centroid coordinate change threshold, and calculating
Obtaining the centroid coordinate difference value omega of the jth same object in the front and back different key frame imagesjThe centroid coordinate of the jth identical object in the (i + 1) th key frame image is (x)i+1，yi+1) The centroid coordinate of the jth identical object in the ith key frame image is (x)i，yi) When the jth same object is in the centroid coordinate difference value omega in the front and back different key frame imagesjAnd when the change time is less than the centroid coordinate change threshold, judging the jth same object as the left-behind object, and acquiring the stationary time length of the jth same object after being left behind.
In a second aspect, the invention provides a detection system for a left article in security monitoring, which comprises a monitoring video frame image acquisition module, a target detection network model establishing module and a left article detection module;
the monitoring video frame image acquisition module is used for acquiring monitoring video frame image datasets in different scenes and acquiring video stream image data of a to-be-detected article to be left, preprocessing the monitoring video frame image datasets, setting a label category for each image in the monitoring video frame image datasets, and acquiring a preprocessed monitoring video frame image dataset and a label dataset corresponding to the preprocessed monitoring video frame image dataset; acquiring a plurality of key frame images in video stream image data according to a time sequence;
the target detection network model establishing module is used for establishing a target detection network model, the target detection network model is a convolutional neural network model added into the SPP module, and the target detection network model is trained according to a monitoring video frame image data set and a label data set which are preprocessed in the monitoring video frame image acquisition module to obtain a trained target detection network model;
the legacy article detection module is used for respectively inputting a plurality of key frame images acquired by the monitoring video frame image acquisition module into the trained target detection network model according to the time sequence, and the trained target detection network model respectively outputs coordinate bounding boxes and labels of all objects in each key frame image; calculating the coordinates of the mass centers of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in front and back different key frame images according to the coordinates of the mass centers and by adopting a front and back frame minimum mass center distance matching method, acquiring the static time of the same object in the matched front and back different key frame images after the object is left, and outputting object left-behind early warning information according to the static time.
By adopting the technical scheme disclosed by the invention, the following beneficial effects are obtained:
compared with the existing image object detection method, the invention provides an object leaving method based on deep learning, which utilizes a computer vision technology to detect objects and pedestrians and provides a detector which is stable for a long time.
Compared with a detection method based on Darknet53, the method disclosed by the invention has the advantages that the recall rate of small articles is improved by performing data enhancement on the image and integrating the SPP module in front of the prediction characteristic diagram.
For the event of article leaving, the invention matches each object of different frames based on the minimum distance of the mass center, the method matches the same object quickly in the boundary frame information output by the detector, the matching accuracy is high, and the early warning information is sent out in time after the article is left.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that it will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the principle of the present invention, and such modifications and improvements should also be considered within the scope of the present invention.
Claims (7)
1. A detection method for a left article in security monitoring is characterized by comprising the following steps:
s1) acquiring monitoring video frame image datasets in different scenes, preprocessing the monitoring video frame image datasets, setting label types for each image in the monitoring video frame image datasets, and acquiring the preprocessed monitoring video frame image datasets and label datasets corresponding to the preprocessed monitoring video frame image datasets;
s2), establishing a target detection network model, wherein the target detection network model is a convolutional neural network model added with an SPP module, and training the target detection network model according to the preprocessed monitoring video frame image data set and the label data set to obtain a trained target detection network model;
s3) collecting video stream image data of a to-be-detected left article, acquiring a plurality of key frame images in the video stream image data according to a time sequence, respectively inputting the plurality of key frame images into the trained target detection network model according to the time sequence, and respectively outputting coordinate bounding boxes and labels of all objects in each key frame image by the trained target detection network model;
s4) calculating the coordinates of the centers of mass of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in the front and back different key frame images according to the coordinates of the centers of mass and by adopting a front and back frame minimum center of mass distance matching method, acquiring the remaining static time length of the same object in the front and back different key frame images, setting a time length threshold value, judging whether the remaining static time length of the same object in the matched front and back different key frame images exceeds the time length threshold value, if not, returning to the step S3); if yes, outputting object leaving early warning information.
2. The method for detecting the left-over article in the security monitoring according to claim 1, wherein in step S1), the preprocessing is performed on the monitoring video frame image data set, and a label category is set for each image in the monitoring video frame image data set, including performing an enhancement preprocessing on each image data in the monitoring video frame image data set, where the enhancement preprocessing includes image cropping, image flipping, and Mosaic data enhancement; the tags include pedestrians, cartons, bags and luggage.
3. The method for detecting the left-over articles in the security protection monitoring according to claim 1 or 2, wherein in the step S2), the convolutional neural network model added with the SPP module comprises an input layer, a middle layer and an output layer; the middle layer comprises a plurality of convolution layers, a plurality of batch normalization layers, a plurality of nonlinear activation function layers and an SPP module; the output layer outputs object coordinate bounding boxes, confidence levels and labels in each image.
4. The method according to claim 3, wherein in step S3), the key frame images in the video stream image data are obtained in time sequence, including setting a preset number of frames, and key frame images are collected in the video stream image data every other preset number of frames in time sequence.
5. The method for detecting the left-behind object in the security monitoring according to claim 4, wherein in step S4), the same object in the previous and subsequent key frame images is matched according to the centroid coordinates and by using the matching method of the minimum centroid distance of the previous and subsequent frames, including matching and associating between different key frame images by using the minimum distance between the centroid coordinates of the object as the feature, and the same object located in different key frame images is obtained through the matching and associating.
6. The method for detecting the left article in the security monitoring according to claim 5, wherein in step S4), the still time lengths of the same object after being left in different keyframe images before and after the object is left are obtained,setting a centroid coordinate change threshold value, and calculating
Obtaining the centroid coordinate difference value omega of the jth same object in the front and back different key frame imagesjThe centroid coordinate of the jth identical object in the (i + 1) th key frame image is (x)i+1，yi+1) The centroid coordinate of the jth identical object in the ith key frame image is (x)i，yi) When the jth same object is in the centroid coordinate difference value omega in the front and back different key frame imagesjAnd when the change time is less than the centroid coordinate change threshold value, judging the jth same object as the left-behind object, and acquiring the stationary time length of the jth same object after the jth same object is left behind.
7. A detection system of a left article in security monitoring is suitable for the detection method of the left article in the security monitoring according to any one of claims 1 to 6, and is characterized by comprising a monitoring video frame image acquisition module, a target detection network model establishment module and a left article detection module;
the monitoring video frame image acquisition module is used for acquiring monitoring video frame image data sets in different scenes and acquiring video stream image data of a to-be-detected article to be left, preprocessing the monitoring video frame image data sets, setting a label category for each image in the monitoring video frame image data sets, and acquiring a preprocessed monitoring video frame image data set and a label data set corresponding to the preprocessed monitoring video frame image data set; acquiring a plurality of key frame images in the video stream image data according to a time sequence;
the target detection network model establishing module is used for establishing a target detection network model, the target detection network model is a convolutional neural network model added into an SPP module, and the target detection network model is trained according to a monitoring video frame image data set and a label data set which are preprocessed in the monitoring video frame image acquisition module to obtain a trained target detection network model;
the legacy article detection module is used for respectively inputting a plurality of key frame images acquired by the monitoring video frame image acquisition module into the trained target detection network model according to the time sequence, and the trained target detection network model respectively outputs coordinate bounding boxes and labels of all objects in each key frame image; calculating the coordinates of the mass centers of all objects in each key frame image according to the coordinate bounding boxes of all objects in each key frame image, matching the same object in the front and back different key frame images according to the coordinates of the mass centers and by adopting a front and back frame minimum mass center distance matching method, acquiring the static time length of the same object in the matched front and back different key frame images after the object is left, and outputting object left-behind early warning information according to the static time length.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011564655.3A CN112699769A (en) | 2020-12-25 | 2020-12-25 | Detection method and system for left-over articles in security monitoring |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011564655.3A CN112699769A (en) | 2020-12-25 | 2020-12-25 | Detection method and system for left-over articles in security monitoring |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112699769A true CN112699769A (en) | 2021-04-23 |
Family
ID=75510692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011564655.3A Pending CN112699769A (en) | 2020-12-25 | 2020-12-25 | Detection method and system for left-over articles in security monitoring |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112699769A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689459A (en) * | 2021-07-30 | 2021-11-23 | 南京信息工程大学 | GMM (Gaussian mixture model) combined with YOLO (YOLO) based real-time tracking and graph building method in dynamic environment |
| CN113763683A (en) * | 2021-09-09 | 2021-12-07 | 南京奥拓电子科技有限公司 | Method and device for reminding article leaving and storage medium |
| CN114973065A (en) * | 2022-04-29 | 2022-08-30 | 北京容联易通信息技术有限公司 | Method and system for detecting article moving and leaving based on video intelligent analysis |
| CN115690046A (en) * | 2022-10-31 | 2023-02-03 | 江苏慧眼数据科技股份有限公司 | Article legacy detection and tracing method and system based on monocular depth estimation |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156942A (en) * | 2014-07-02 | 2014-11-19 | 华南理工大学 | Detection method for remnants in complex environment |
| CN107527009A (en) * | 2017-07-11 | 2017-12-29 | 浙江汉凡软件科技有限公司 | A kind of remnant object detection method based on YOLO target detections |
-
2020
- 2020-12-25 CN CN202011564655.3A patent/CN112699769A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156942A (en) * | 2014-07-02 | 2014-11-19 | 华南理工大学 | Detection method for remnants in complex environment |
| CN107527009A (en) * | 2017-07-11 | 2017-12-29 | 浙江汉凡软件科技有限公司 | A kind of remnant object detection method based on YOLO target detections |
Non-Patent Citations (1)
| Title |
|---|
| 刘冰 等: "《深度核机器学习技术及应用》", 30 November 2019, 北京工业大学出版社, pages: 134 - 143 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689459A (en) * | 2021-07-30 | 2021-11-23 | 南京信息工程大学 | GMM (Gaussian mixture model) combined with YOLO (YOLO) based real-time tracking and graph building method in dynamic environment |
| CN113689459B (en) * | 2021-07-30 | 2023-07-18 | 南京信息工程大学 | Real-time tracking and mapping method based on GMM and YOLO under dynamic environment |
| CN113763683A (en) * | 2021-09-09 | 2021-12-07 | 南京奥拓电子科技有限公司 | Method and device for reminding article leaving and storage medium |
| CN114973065A (en) * | 2022-04-29 | 2022-08-30 | 北京容联易通信息技术有限公司 | Method and system for detecting article moving and leaving based on video intelligent analysis |
| CN115690046A (en) * | 2022-10-31 | 2023-02-03 | 江苏慧眼数据科技股份有限公司 | Article legacy detection and tracing method and system based on monocular depth estimation |
| CN115690046B (en) * | 2022-10-31 | 2024-02-23 | 江苏慧眼数据科技股份有限公司 | Article carry-over detection and tracing method and system based on monocular depth estimation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112699769A (en) | Detection method and system for left-over articles in security monitoring | |
| Benezeth et al. | Abnormal events detection based on spatio-temporal co-occurences | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| CN110852179B (en) | Suspicious personnel invasion detection method based on video monitoring platform | |
| Mathur et al. | Research on intelligent video surveillance techniques for suspicious activity detection critical review | |
| Chen et al. | Vehicles detection on expressway via deep learning: Single shot multibox object detector | |
| Park et al. | Study on 3D action recognition based on deep neural network | |
| Karunakaran | Deep learning based object detection using mask RCNN | |
| Kalaivani et al. | Towards comprehensive understanding of event detection and video summarization approaches | |
| CN116310933A (en) | Cross-mirror tracking early warning method and system for rail transit intrusion scene personnel | |
| Sadkhan et al. | An investigate on moving object tracking and detection in images | |
| Hassan et al. | Crowd counting using deep learning based head detection | |
| Sathesh et al. | Speedy detection module for abandoned belongings in airport using improved image processing technique | |
| Sridhar et al. | Anomaly Detection using CNN with SVM | |
| Lee et al. | Recognizing human-vehicle interactions from aerial video without training | |
| Ravikumar et al. | Video Surveillance for Intruder Detection and Tracking | |
| Bhardwaj et al. | Modified Neural Network-based Object Classification in Video Surveillance System. | |
| Park et al. | Video surveillance system based on 3d action recognition | |
| Gautam et al. | Suspicious Object Tracking With Yolov3 with Python Using Open-CV | |
| Xu et al. | An intra-frame classification network for video anomaly detection and localization | |
| Lahouli et al. | A fast and robust approach for human detection in thermal imagery for surveillance using UAVs | |
| Prabha et al. | Human anomalous activity detection: shape and motion approach in crowded scenes | |
| Kathole et al. | License Plate Recognition for Detecting Stolen Vehicle Using Deep Learning | |
| Babaryka et al. | Deep Learning Methods Application for Object Detection Tasks Using Unmanned Aerial Vehicles | |
| CN118262288B (en) | System, method and medium for tracking and monitoring suspicious personnel in large-scale stadium |
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 |