CN112668432A - Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort - Google Patents
Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort Download PDFInfo
- Publication number
- CN112668432A CN112668432A CN202011531976.3A CN202011531976A CN112668432A CN 112668432 A CN112668432 A CN 112668432A CN 202011531976 A CN202011531976 A CN 202011531976A CN 112668432 A CN112668432 A CN 112668432A
- Authority
- CN
- China
- Prior art keywords
- human body
- tracking
- target
- yolov5
- body detection
- 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 64
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 4
- 230000003993 interaction Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a human body detection tracking method in a ground interactive projection system based on YoloV5 and deep sort, which comprises the following steps: step 1: acquiring an image to be identified, and preprocessing the image; step 2: constructing a human body detection tracking model based on YoloV5 and Deepsort; and step 3: training a human body detection tracking model; and 4, step 4: calculating the accuracy of the model, judging whether the accuracy is greater than a preset threshold value, if so, executing the step 5, otherwise, returning to the step 3; and 5: and (3) detecting and tracking the human body in the ground interactive projection system in real time by using the trained human body detection and tracking model. Compared with the prior art, the method has the advantages of high precision, good real-time performance and the like.
Description
Technical Field
The invention relates to the technical field of ground interactive human body detection and tracking, in particular to a human body detection and tracking method in a ground interactive projection system based on YoloV5 and deep sort.
Background
The ground interactive projection can change any one flat and light floor area into a cool and high-tech real-time interactive magic ground. The real interaction between the participants and the images on the ground enables a plurality of participants to be blended into the scene and participate in the game at the same time. Rich content can be displayed, and various interesting interactive mini-games can be contained. Spectators can interact with the content in the projection picture with limbs, bring an unusual advertisement and the effect that the amusement is shone and is shone mutually, can play fine active atmosphere, increase the scientific and technological content of show, also provide the bandwagon effect that has creative content for enterprise's user, improved on-the-spot people's strength and relevant unit's image, have fine propaganda effect.
Human interaction is the key of the ground interactive projection system, and if the projection system wants to interact with a user, the human body needs to be detected and tracked, so that the interaction with the human body can be further realized. At present, the algorithm for detecting and tracking human body on the ground interactive projection is not studied much. Chinese patent CN110516556A discloses a multi-target tracking detection method based on Darkflow-Deepsort, which realizes the detection and tracking of multiple targets by using YoloV3 and Deepsort algorithms, but the target detection precision and the tracking precision of the algorithms are low, and the running speed of the algorithms is low, so that the requirements of a ground interactive projection system are not met.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a human body detection and tracking method based on YooloV 5 and Deepsort in a ground interactive projection system, which has high precision and good real-time performance.
The purpose of the invention can be realized by the following technical scheme:
a human body detection tracking method in a ground interactive projection system based on YoloV5 and Deepsort comprises the following steps:
step 1: acquiring an image to be identified, and preprocessing the image;
step 2: constructing a human body detection tracking model based on YoloV5 and Deepsort;
and step 3: training a human body detection tracking model;
and 4, step 4: calculating the accuracy of the model, judging whether the accuracy is greater than a preset threshold value, if so, executing the step 5, otherwise, returning to the step 3;
and 5: and (3) detecting and tracking the human body in the ground interactive projection system in real time by using the trained human body detection and tracking model.
Preferably, the step 1 specifically comprises:
acquiring a set of images to be identified, randomly dividing the images into a training set and a testing set, performing Mosaic data enhancement, self-adaptive anchor frame calculation and self-adaptive picture scaling on all the images, and expanding the training set and the testing set.
Preferably, the human body detection tracking model in step 2 is specifically:
step 2-1: building a YoloV5 submodel for detecting a human body;
step 2-2: and constructing a deep sort submodel for tracking the detected human body.
More preferably, the YoloV5 submodel is specifically:
the YoloV5 submodel comprises a Backbone network of Backbone, a Neck network and a Head network which are connected in sequence;
the Backbone network of the backhaul is a network formed by combining a Focus structure and a CSP structure and is used for acquiring characteristic images of the image pairs;
the Neck network is a network formed by combining an FPN structure and a PAN structure, is used for mixing and combining the characteristics of the images and transmitting the characteristic images to a prediction layer;
the Head network is used for predicting the image characteristics, obtaining a boundary frame and predicting the category; the Head network uses GIOU _ Loss as a Loss function of a Bounding box; the anchor boxes are screened using a weighted NMS.
More preferably, the DeepSort submodel specifically comprises:
firstly, the state of the detected human body target frame is estimated, the state of the target is predicted, and then the detection result and the prediction result are associated and matched.
More preferably, the state estimation method specifically includes:
using an eight-dimensional spaceDescribing the state of the trajectory at a certain time, wherein (u, v) is the center coordinates of the target frame, r is the aspect ratio of the target frame, h is the height of the target frame,corresponding speed information in an image coordinate system;
predicting the motion state of the target by using a standard Kalman filter based on a constant velocity model and a linear observation model, wherein the predicted result is (u, v, r, h);
for each tracking target, recording the frame number alpha after the last detection result is matched with the tracking resultkSetting the parameter to 0 once the monitoring result of one target is correctly associated with the tracking result; if α iskExceeds the set maximum threshold AmaxIf the target is not tracked, determining that the tracking process of the target is finished;
the judgment of the new target is: if the prediction result of the target position of the potential new tracker in three continuous frames can be correctly associated with the detection result, the new moving target is confirmed to be present; if the requirement cannot be met, the moving object needs to be deleted.
More preferably, the association matching method includes association matching of motion informationMatching and appearance information correlation matching; associating a degree of match d using motion information(1)(i, j) and appearance information association matching degree d(2)(i, j) as the final correlation matching degree ci,jThe method specifically comprises the following steps:
using ci,jWeight per line, usingDetermining the connecting line of the initial match, bijIs a value of 0/1, bijWhen 1, it means that there is a connection between i and j, bij0 means that there is no connection between i and j;
ci,jthe calculation method comprises the following steps:
ci,j=λd(1)(i,j)+(1-λ)d(2)(i,j)
wherein λ is a hyper-parameter.
More preferably, the motion information association matching specifically includes:
measuring the distance between the Track and the prediction result Detection by using the squared Mahalanobis distance, wherein the calculation method comprises the following steps:
wherein d isjIs the jth prediction result Detection; y isiIs the ith Track;covariance as d and y;is an indicator that if the mahalanobis distance is less than the threshold t(1)Representing a successful match.
More preferably, the appearance information association matching specifically includes:
for each prediction result djSolving for a unit norm eigenvector ri;
Calculating the latest 100 successful association feature sets of the ith track and the jth prediction result d of the current framejFeature vector riMinimum cosine distance between:
is an indicator if the minimum cosine distance is less than a threshold value t(1)Representing a successful match.
More preferably, the association matching method further includes cascade matching, specifically:
when a target is shielded for a long time, the uncertainty of Kalman filtering prediction is greatly increased, if two Kalman filters compete for the matching right of the same detection result at the same time, the detection result is associated with a track with longer shielding time, and at the moment, cascade matching is introduced to give priority to the target which appears more frequently.
Compared with the prior art, the invention has the following beneficial effects:
firstly, the human body detection and tracking precision is high: according to the human body detection tracking method in the ground interactive projection system, the YoloV5 model and the Deepsort model are combined, the YoloV5 model is used for detecting the human body of the ground interactive system, and the Deepsort model is used for tracking the detected human body, so that the human body detection and tracking precision is greatly improved, and a foundation is provided for realizing the ground interactive projection system.
Secondly, the algorithm is fast in speed and good in real-time performance: the image processing speed of the YoloV5 human body detection model used by the human body detection and tracking method in the ground interactive projection system is more than 100 times of the image processing speed of the human body detection model in the prior art, so that the real-time performance of human body detection and tracking of the ground interactive projection system is greatly improved.
Drawings
FIG. 1 is a schematic flow chart of a human body detection and tracking method in a ground interactive projection system according to the present invention;
fig. 2 is a schematic structural diagram of YoloV5 model in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
A human body detection tracking method in a floor interactive projection system based on yoolov 5 and deep sort, a flow of which is shown in fig. 1, includes:
step 1: acquiring an image to be identified, and preprocessing the image, specifically:
randomly dividing the images into a training set and a test set, performing Mosaic data enhancement, adaptive anchor frame calculation and adaptive picture scaling on all the images, and expanding the training set and the test set;
(1) mosaic data enhancement
The method comprises the steps of splicing in the modes of random zooming, random cutting and random arrangement, and has a good detection effect on small targets;
(2) adaptive anchor frame computing
For different data sets, there will be an anchor frame with the length and width initially set. In the network training, the network outputs a prediction frame on the basis of an initial anchor frame, and then compares the prediction frame with a real frame ground route, calculates the difference between the prediction frame and the real frame ground route, and then reversely updates and iterates network parameters.
(3) Adaptive picture scaling
Since the length and width of different pictures are different, the common method is to scale the original picture to a standard size and send the scaled original picture to the detection network. For example, dimensions 416 × 416, 608 × 608 are commonly used in the Yolo algorithm. In actual use, the aspect ratios of many pictures are different, so after the zoom filling, the sizes of the black edges at two ends are different, and if the filling is more, information redundancy exists, which affects the reasoning speed. Thus, the letterbox function of datasets. py in the code of Yolov5 is modified to adaptively add the least black edges to the original image. The black edges at the two ends of the image height are reduced, and the calculated amount is reduced during reasoning, namely the target detection speed is improved. The reasoning speed is improved by 37%, so that the effect is obvious.
Step 2: constructing a human body detection tracking model based on YoloV5 and Deepsort, which specifically comprises the following steps:
step 2-1: a YoloV5 submodel is set up for detecting human bodies, and the structure of the YoloV5 submodel is shown in figure 2;
the YoloV5 submodel comprises a Backbone network of Backbone, a Neck network and a Head network which are connected in sequence;
the Backbone network of the backhaul is a network formed by combining a Focus structure and a CSP structure and is used for acquiring characteristic images of the image pairs;
the Neck network is a network formed by combining an FPN structure and a PAN structure, is used for mixing and combining the characteristics of the images and transmitting the characteristic images to a prediction layer;
the Head network is connected with the output end and used for predicting the image characteristics to obtain a boundary frame and predict the category; the Head network uses GIOU _ Loss as a Loss function of a Bounding box; screening the anchor frame by using a weighted NMS mode;
step 2-2: building a deep sort submodel for tracking the detected human body;
the deep sort submodel specifically comprises: firstly, carrying out state estimation on a detected human body target frame, predicting the state of a target, and then carrying out correlation matching on a detection result and a prediction result;
the state estimation method specifically comprises the following steps:
using an eight-dimensional spaceDescribing the state of the trajectory at a certain time, wherein (u, v) is the center coordinates of the target frame, r is the aspect ratio of the target frame, h is the height of the target frame,corresponding speed information in an image coordinate system;
predicting the motion state of the target by using a standard Kalman filter based on a constant velocity model and a linear observation model, wherein the predicted result is (u, v, r, h);
for each tracking target, recording the frame number alpha after the last detection result is matched with the tracking resultkSetting the parameter to 0 once the monitoring result of one target is correctly associated with the tracking result; if α iskExceeds the set maximum threshold AmaxIf the target is not tracked, determining that the tracking process of the target is finished;
the judgment of the new target is: if the prediction result of the target position of the potential new tracker in three continuous frames can be correctly associated with the detection result, the new moving target is confirmed to be present; if the requirement can not be met, the moving target needs to be deleted;
the association matching method comprises the steps of association matching of motion information and association matching of appearance information; associating a degree of match d using motion information(1)(i, j) and appearance information association matching degree d(2)(i, j) as the final correlation matching degree ci,jThe method specifically comprises the following steps:
using ci,jWeight per line, usingDetermining the connecting line of the initial match, bijIs a value of 0/1, bijWhen 1, it means that there is a connection between i and j, bij0 means that there is no connection between i and j;
ci,j=λd(1)(i,j)+(1-λ)d(2)(i,j)
8. the method as claimed in claim 7, wherein the motion information association matching specifically comprises:
measuring the distance between the Track and the prediction result Detection by using the squared Mahalanobis distance, wherein the calculation method comprises the following steps:
wherein d isjIs the jth prediction result Detection; y isiIs the ith Track;covariance as d and y;is an indicator that if the mahalanobis distance is less than the threshold t(1)Represents a successful match;
the appearance information association matching specifically comprises the following steps:
for each prediction result djSolving for a unit norm eigenvector ri;
Calculating the latest 100 successful association feature sets of the ith track and the jth prediction result d of the current framejFeature vector riMinimum cosine distance between:
is an indicator if the minimum cosine distance is less than a threshold value t(1)Representing a successful match.
The association matching method in this embodiment further includes cascade matching, and the application timing specifically is:
when a target is shielded for a long time, the uncertainty of Kalman filtering prediction is greatly increased, if two Kalman filters simultaneously compete for the matching right of the same detection result at the moment, the detection result is associated with a track with longer shielding time, and at the moment, the introduction of cascade matching gives priority matching right to the target which appears more frequently;
and step 3: training a human body detection tracking model;
and 4, step 4: calculating the accuracy of the model, judging whether the accuracy is greater than a preset threshold value, if so, executing the step 5, otherwise, returning to the step 3;
and 5: and (3) detecting and tracking the human body in the ground interactive projection system in real time by using the trained human body detection and tracking model.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A human body detection tracking method in a floor interactive projection system based on YoloV5 and Deepsort is characterized in that the detection tracking method comprises the following steps:
step 1: acquiring an image to be identified, and preprocessing the image;
step 2: constructing a human body detection tracking model based on YoloV5 and Deepsort;
and step 3: training a human body detection tracking model;
and 4, step 4: calculating the accuracy of the model, judging whether the accuracy is greater than a preset threshold value, if so, executing the step 5, otherwise, returning to the step 3;
and 5: and (3) detecting and tracking the human body in the ground interactive projection system in real time by using the trained human body detection and tracking model.
2. The method for detecting and tracking the human body in the floor interactive projection system based on YoloV5 and deep sort as claimed in claim 1, wherein the step 1 specifically comprises:
acquiring a set of images to be identified, randomly dividing the images into a training set and a testing set, performing Mosaic data enhancement, self-adaptive anchor frame calculation and self-adaptive picture scaling on all the images, and expanding the training set and the testing set.
3. The method as claimed in claim 1, wherein the human body detection tracking model in the step 2 is specifically:
step 2-1: building a YoloV5 submodel for detecting a human body;
step 2-2: and constructing a deep sort submodel for tracking the detected human body.
4. The human body detection and tracking method in the floor interactive projection system based on yoolov 5 and deep sort as claimed in claim 3, wherein the yoolov 5 submodel is specifically:
the YoloV5 submodel comprises a Backbone network of Backbone, a Neck network and a Head network which are connected in sequence;
the Backbone network of the backhaul is a network formed by combining a Focus structure and a CSP structure and is used for acquiring characteristic images of the image pairs;
the Neck network is a network formed by combining an FPN structure and a PAN structure, is used for mixing and combining the characteristics of the images and transmitting the characteristic images to a prediction layer;
the Head network is used for predicting the image characteristics, obtaining a boundary frame and predicting the category; the Head network uses GIOU _ Loss as a Loss function of a Bounding box; the anchor boxes are screened using a weighted NMS.
5. The method as claimed in claim 3, wherein the deep sort submodel specifically comprises:
firstly, the state of the detected human body target frame is estimated, the state of the target is predicted, and then the detection result and the prediction result are associated and matched.
6. The human body detection tracking method in the floor interactive projection system based on YoloV5 and deep sort as claimed in claim 5, wherein the state estimation method specifically comprises:
using an eight-dimensional spaceDescribing the state of the trajectory at a certain time, wherein (u, v) is the center coordinates of the target frame, r is the aspect ratio of the target frame, h is the height of the target frame,corresponding speed information in an image coordinate system;
predicting the motion state of the target by using a standard Kalman filter based on a constant velocity model and a linear observation model, wherein the predicted result is (u, v, r, h);
for each tracking target, recording the frame number alpha after the last detection result is matched with the tracking resultkSetting the parameter to 0 once the monitoring result of one target is correctly associated with the tracking result; if α iskExceeds the set maximum threshold AmaxIf the target is not tracked, determining that the tracking process of the target is finished;
the judgment of the new target is: if the prediction result of the target position of the potential new tracker in three continuous frames can be correctly associated with the detection result, the new moving target is confirmed to be present; if the requirement cannot be met, the moving object needs to be deleted.
7. The human body detection and tracking method in the YoloV5 and deep sort-based ground interactive projection system according to claim 5, wherein the correlation matching method comprises motion information correlation matching and appearance information correlation matching; associating a degree of match d using motion information(1)(i, j) and appearance information association matching degree d(2)(i, j) as the final correlation matching degree ci,jThe method specifically comprises the following steps:
using ci,jWeight per line, usingDetermining the connecting line of the initial match, bijIs a value of 0/1, bijWhen 1, it means that there is a connection between i and j, bij0 means that there is no connection between i and j;
ci,jthe calculation method comprises the following steps:
ci,j=λd(1)(i,j)+(1-λ)d(2)(i,j)
wherein λ is a hyper-parameter.
8. The method as claimed in claim 7, wherein the motion information association matching specifically comprises:
measuring the distance between the Track and the prediction result Detection by using the squared Mahalanobis distance, wherein the calculation method comprises the following steps:
9. The human body detection and tracking method in the floor interactive projection system based on YoloV5 and deep sort as claimed in claim 7, wherein the appearance information association matching specifically comprises:
for each prediction result djSolving for a unit norm eigenvector ri;
Calculating the latest 100 successful association feature sets of the ith track and the jth prediction result d of the current framejFeature vector riMinimum cosine distance between:
10. The human body detection and tracking method in the floor interactive projection system based on YoloV5 and deep sort as claimed in claim 7, wherein the association matching method further comprises cascade matching, specifically:
when a target is shielded for a long time, the uncertainty of Kalman filtering prediction is greatly increased, if two Kalman filters compete for the matching right of the same detection result at the same time, the detection result is associated with a track with longer shielding time, and at the moment, cascade matching is introduced to give priority to the target which appears more frequently.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011531976.3A CN112668432A (en) | 2020-12-22 | 2020-12-22 | Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011531976.3A CN112668432A (en) | 2020-12-22 | 2020-12-22 | Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112668432A true CN112668432A (en) | 2021-04-16 |
Family
ID=75407824
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011531976.3A Pending CN112668432A (en) | 2020-12-22 | 2020-12-22 | Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112668432A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113158995A (en) * | 2021-05-21 | 2021-07-23 | 西安建筑科技大学 | Multi-target tracking detection method, system, equipment and storage medium |
CN113160062A (en) * | 2021-05-25 | 2021-07-23 | 烟台艾睿光电科技有限公司 | Infrared image target detection method, device, equipment and storage medium |
CN113191204A (en) * | 2021-04-07 | 2021-07-30 | 华中科技大学 | Multi-scale blocking pedestrian detection method and system |
CN113269073A (en) * | 2021-05-19 | 2021-08-17 | 青岛科技大学 | Ship multi-target tracking method based on YOLO V5 algorithm |
CN113298130A (en) * | 2021-05-14 | 2021-08-24 | 嘉洋智慧安全生产科技发展(北京)有限公司 | Target image detection method and target object detection model generation method |
CN113449806A (en) * | 2021-07-12 | 2021-09-28 | 苏州大学 | Two-stage forestry pest identification and detection system and method based on hierarchical structure |
CN113724153A (en) * | 2021-08-02 | 2021-11-30 | 电子科技大学 | Method for eliminating redundant images based on machine learning |
CN114332749A (en) * | 2021-11-16 | 2022-04-12 | 河海大学 | Sand production monitoring method of sand production ship based on deep learning |
CN114494350A (en) * | 2022-01-28 | 2022-05-13 | 北京中电兴发科技有限公司 | Personnel gathering detection method and device |
CN115375736A (en) * | 2022-10-25 | 2022-11-22 | 威海市博华医疗设备有限公司 | Image-based pedestrian trajectory tracking method and device |
CN115723152A (en) * | 2022-11-17 | 2023-03-03 | 中国人民解放军总医院第五医学中心 | Intelligent nursing robot |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109871763A (en) * | 2019-01-16 | 2019-06-11 | 清华大学 | A kind of specific objective tracking based on YOLO |
CN110490901A (en) * | 2019-07-15 | 2019-11-22 | 武汉大学 | The pedestrian detection tracking of anti-attitudes vibration |
CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技(深圳)有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
CN111144247A (en) * | 2019-12-16 | 2020-05-12 | 浙江大学 | Escalator passenger reverse-running detection method based on deep learning |
WO2020155873A1 (en) * | 2019-02-02 | 2020-08-06 | 福州大学 | Deep apparent features and adaptive aggregation network-based multi-face tracking method |
CN111860282A (en) * | 2020-07-15 | 2020-10-30 | 中国电子科技集团公司第三十八研究所 | Subway section passenger flow volume statistics and pedestrian retrograde motion detection method and system |
CN111914664A (en) * | 2020-07-06 | 2020-11-10 | 同济大学 | Vehicle multi-target detection and track tracking method based on re-identification |
CN112101433A (en) * | 2020-09-04 | 2020-12-18 | 东南大学 | Automatic lane-dividing vehicle counting method based on YOLO V4 and DeepsORT |
-
2020
- 2020-12-22 CN CN202011531976.3A patent/CN112668432A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109871763A (en) * | 2019-01-16 | 2019-06-11 | 清华大学 | A kind of specific objective tracking based on YOLO |
WO2020155873A1 (en) * | 2019-02-02 | 2020-08-06 | 福州大学 | Deep apparent features and adaptive aggregation network-based multi-face tracking method |
CN110490901A (en) * | 2019-07-15 | 2019-11-22 | 武汉大学 | The pedestrian detection tracking of anti-attitudes vibration |
CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技(深圳)有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
CN111144247A (en) * | 2019-12-16 | 2020-05-12 | 浙江大学 | Escalator passenger reverse-running detection method based on deep learning |
CN111914664A (en) * | 2020-07-06 | 2020-11-10 | 同济大学 | Vehicle multi-target detection and track tracking method based on re-identification |
CN111860282A (en) * | 2020-07-15 | 2020-10-30 | 中国电子科技集团公司第三十八研究所 | Subway section passenger flow volume statistics and pedestrian retrograde motion detection method and system |
CN112101433A (en) * | 2020-09-04 | 2020-12-18 | 东南大学 | Automatic lane-dividing vehicle counting method based on YOLO V4 and DeepsORT |
Non-Patent Citations (1)
Title |
---|
新智元: "深入浅出Yolo系列之Yolov5核心基础知识完整讲解", 知乎, pages 1 - 18 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113191204A (en) * | 2021-04-07 | 2021-07-30 | 华中科技大学 | Multi-scale blocking pedestrian detection method and system |
CN113191204B (en) * | 2021-04-07 | 2022-06-17 | 华中科技大学 | Multi-scale blocking pedestrian detection method and system |
CN113298130A (en) * | 2021-05-14 | 2021-08-24 | 嘉洋智慧安全生产科技发展(北京)有限公司 | Target image detection method and target object detection model generation method |
CN113269073A (en) * | 2021-05-19 | 2021-08-17 | 青岛科技大学 | Ship multi-target tracking method based on YOLO V5 algorithm |
CN113158995A (en) * | 2021-05-21 | 2021-07-23 | 西安建筑科技大学 | Multi-target tracking detection method, system, equipment and storage medium |
CN113160062A (en) * | 2021-05-25 | 2021-07-23 | 烟台艾睿光电科技有限公司 | Infrared image target detection method, device, equipment and storage medium |
CN113449806A (en) * | 2021-07-12 | 2021-09-28 | 苏州大学 | Two-stage forestry pest identification and detection system and method based on hierarchical structure |
CN113724153A (en) * | 2021-08-02 | 2021-11-30 | 电子科技大学 | Method for eliminating redundant images based on machine learning |
CN114332749B (en) * | 2021-11-16 | 2024-04-19 | 河海大学 | Sand production monitoring method of sand production ship based on deep learning |
CN114332749A (en) * | 2021-11-16 | 2022-04-12 | 河海大学 | Sand production monitoring method of sand production ship based on deep learning |
CN114494350A (en) * | 2022-01-28 | 2022-05-13 | 北京中电兴发科技有限公司 | Personnel gathering detection method and device |
CN114494350B (en) * | 2022-01-28 | 2022-10-14 | 北京中电兴发科技有限公司 | Personnel gathering detection method and device |
CN115375736A (en) * | 2022-10-25 | 2022-11-22 | 威海市博华医疗设备有限公司 | Image-based pedestrian trajectory tracking method and device |
CN115723152A (en) * | 2022-11-17 | 2023-03-03 | 中国人民解放军总医院第五医学中心 | Intelligent nursing robot |
CN115723152B (en) * | 2022-11-17 | 2023-06-06 | 中国人民解放军总医院第五医学中心 | Intelligent nursing robot |
WO2024103733A1 (en) * | 2022-11-17 | 2024-05-23 | 中国人民解放军总医院第五医学中心 | Intelligent care robot |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112668432A (en) | Human body detection tracking method in ground interactive projection system based on YoloV5 and Deepsort | |
CN103369209B (en) | Vedio noise reduction device and method | |
CN110378997B (en) | ORB-SLAM 2-based dynamic scene mapping and positioning method | |
JP5487298B2 (en) | 3D image generation | |
CN108182690B (en) | A kind of infrared Weak target detecting method based on prospect weighting local contrast | |
CN104794737B (en) | A kind of depth information Auxiliary Particle Filter tracking | |
CN111126278B (en) | Method for optimizing and accelerating target detection model for few-class scene | |
JP7292492B2 (en) | Object tracking method and device, storage medium and computer program | |
CN109063549A (en) | High-resolution based on deep neural network is taken photo by plane video moving object detection method | |
Zhang et al. | Object detection/tracking toward underwater photographs by remotely operated vehicles (ROVs) | |
CN112906475B (en) | Artificial intelligence-based rolling shutter imaging method and system for urban surveying and mapping unmanned aerial vehicle | |
CN112507848B (en) | Mobile terminal real-time human face attitude estimation method | |
CN106022266A (en) | Target tracking method and target tracking apparatus | |
WO2024060978A1 (en) | Key point detection model training method and apparatus and virtual character driving method and apparatus | |
Chen et al. | Salbinet360: Saliency prediction on 360 images with local-global bifurcated deep network | |
CN115100744A (en) | Badminton game human body posture estimation and ball path tracking method | |
Li et al. | Weak moving object detection in optical remote sensing video with motion-drive fusion network | |
CN108010058A (en) | A kind of method and system that vision tracking is carried out to destination object in video flowing | |
JP7290546B2 (en) | 3D model generation apparatus and method | |
CN110502968B (en) | Method for detecting infrared small and weak moving target based on track point space-time consistency | |
CN111339934A (en) | Human head detection method integrating image preprocessing and deep learning target detection | |
CN113763432B (en) | Target detection tracking method based on image definition and tracking stability conditions | |
CN109905565A (en) | Video stabilization method based on motor pattern separation | |
Nickel et al. | Dynamic integration of generalized cues for person tracking | |
CN115512263A (en) | Dynamic visual monitoring method and device for falling object |
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 |