CN109711322A - A kind of people's vehicle separation method based on RFCN - Google Patents
A kind of people's vehicle separation method based on RFCN Download PDFInfo
- Publication number
- CN109711322A CN109711322A CN201811585378.7A CN201811585378A CN109711322A CN 109711322 A CN109711322 A CN 109711322A CN 201811585378 A CN201811585378 A CN 201811585378A CN 109711322 A CN109711322 A CN 109711322A
- Authority
- CN
- China
- Prior art keywords
- target
- people
- rfcn
- vehicle
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000926 separation method Methods 0.000 title claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 28
- 238000012360 testing method Methods 0.000 claims abstract description 20
- 238000013135 deep learning Methods 0.000 claims abstract description 10
- 230000003068 static effect Effects 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 32
- 230000009467 reduction Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000004321 preservation Methods 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 8
- 238000012544 monitoring process Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The present invention provides a kind of people's vehicle separation method based on RFCN, obtains the video frame images of monitored picture directly from camera including computer, is simply pre-processed to image；Whole image is detected using preparatory trained deep learning detection model, all people, vehicle target in image are identified and positioned, boundary rectangle, classification and the confidence level of target are exported；Computer post-processes above-mentioned testing result, and the removal boundary rectangle frame that confidence level is low, degree of overlapping is high is exported as final result；The continuous video frame images of comprehensive analysis, judge the motion state of target, the reference conditions into the time of scene, as judgement triggering warning.People's vehicle separation method of the present invention based on RFCN can accurately realize that the positioning of the judgement and position to target category returns, and the target that not only can detecte movement also can detecte static target, can accurately filter out inhuman, vehicle target.
Description
Technical field
The invention belongs to field of video monitoring, more particularly, to a kind of people's vehicle separation method based on RFCN.
Background technique
Along with the development and social progress of new and high technology, video monitoring system obtains extensively in more and more occasions
Application.The camera vigilance performance of monitoring system is unable to satisfy a variety of different complex scenes and more and more differences and needs at present
It asks.Be mainly reflected in, warning scene requirement monitoring system can not only accurate detection go out the target in scene, and can also from
The position of the dynamic classification for identifying target and place.Main monitoring objective under daily warning scene is people and Che, so needing
The target category of detection is also people or vehicle, it is therefore desirable to which a kind of method is accurate to the target progress of detection to identify classification.It is existing
Some warning cameras can be substantially real using the detection such as cascade detection, support vector machines or simple BP neural network classification method
Now to the classification arbitration functions of detection target, but the scene that these methods adapt to is limited, and to the standard of target category detection
True rate is lower, is even more especially deficiency to the identification of Small object and classifying quality, is located in advance furthermore with background modeling technology
The method of reason can only detect the target of movement, be unable to satisfy the demand of static target in detection scene, it is difficult to wide popularization and application.
Summary of the invention
In view of this, the present invention is directed to propose a kind of people's vehicle separation method based on RFCN, for realizing monitoring scene
Under vigilance performance, realization all targets in monitoring scene are fast and accurately positioned and are identified, facilitate relevant people
The target category of member's analysis triggering alarm and the motion state of target.
In order to achieve the above objectives, the technical scheme of the present invention is realized as follows:
A kind of people's vehicle separation method based on RFCN includes the following steps: that 1) computer obtains monitoring directly from camera
The video frame images of picture simply pre-process image；2) preparatory trained deep learning detection model pair is utilized
Whole image is detected, and is identified and positioned to all people, vehicle target in image, is exported boundary rectangle, the class of target
Other and confidence level；3) computer post-processes the testing result in 2), and it is external that removal confidence level is low, degree of overlapping is high
Rectangle frame is exported as final result；4) computer generalization analyzes continuous video frame images, judges the movement shape of target
State, the reference conditions into the time of scene, as judgement triggering warning.
Further, the specific method of the step 1) is handled image noise reduction including the use of Mean Filtering Algorithm, if
More fuzzy picture suitably does acutance processing at night, then by whole image inspection.
Further, the method detected in the step 2) to whole image includes:
201) people, the vehicle sample set of training pattern are constructed；
202) the deep learning network architecture based on RFCN is built；
203) related training parameter, training pattern are configured；
204) detection and the output test result to people, vehicle are realized with training pattern.
Further, the people of building training pattern in the step 201), the specific method of vehicle training sample set include: from
It is obtained in a large amount of actual monitored scene in a variety of actual scenes such as video frame images, including different angle, different light conditions
People, vehicle image with triggering warning alert if.Data enhancing, including rotation, mirror image, sanction are carried out at random to original image
Cut, histogram equalization, gray proces, change brightness, saturation degree, the different enhancement methods such as contrast random combine, thus
Generate sample set abundant.Then marked in the picture using annotation tool people, vehicle boundary rectangle frame and set generic,
Image and corresponding mark file are saved and are used as sample set.
Further, it includes: to adopt that the method based on the RFCN building deep learning network architecture is built in the step 202)
It uses ResNet-18 network as the basic network of image characteristics extraction, the characteristic pattern that basic network exports is input to RFCN mesh
Mark detection framework, the final detection and identification realized to target；Wherein the size and ratio of anchors should meet sample set as far as possible
The scale and proportional region of middle target utilize the position (x, y, w, h) of prediction block, confidence level confidence, true convenient for network
The position of real value and classification carry out costing bio disturbance, calculate the loss reduction so that final by iterating.
Further, it is 0.001 that configuration training parameter, which is mainly the initial learning rate of setting network, in the step 203),
And every to carry out total four times of training samples number of the number of iterations, learning rate is reduced to original 0.5 times, and total the number of iterations is generally set
It is set to ten times of total number of samples amount.Model training is iterated instruction according to backpropagation thought, using stochastic gradient descent algorithm
Practice, the loss value of network is made to be down to lower value.
Further, the detection to people, vehicle is realized with training pattern in the step 204) and output test result is main
Image to be detected is detected including the use of preparatory trained model, the testing result of model output includes the external of target
Classification and corresponding confidence level belonging to the coordinate of rectangle frame, target.
Further, the method post-processed in the step 3) to testing result mainly includes according to each target
A threshold value is arranged in the value of the confidence, removes the lower target of the value of the confidence, and an IOU parameter is arranged, and removes the higher target of degree of overlapping
The value of the confidence is lower in frame, and what these frames may detect is the same target, only retains a highest detection block of the value of the confidence and is
It can.According to the size of detection block, removal does not meet the testing result of target sizes range in actual scene.
Further, the continuous video frame images of comprehensive analysis in the step 4), mainly according to the detection of successive frame
As a result the position of the same target exported in judges that the state of the target is static or movement, and can be to the movement of target
Track is tracked.According to detecting that the video frame number of same target judges that target enters the time of scene, triggered as judgement
The reference conditions of warning.
Compared with the existing technology, people's vehicle separation method of the present invention based on RFCN has the advantage that
(1) people's vehicle separation method of the present invention based on RFCN carries out feature extraction using ResNet18 network, adopts
It uses RFCN as detection framework, may be implemented to carry out target in image accurate kind judging and position returns, to small mesh
Target detection effect is also preferable, and model has stronger generalization ability, can satisfy a variety of different scene demands.Model
Directly whole image is detected, the target that not only can detecte movement can also detect static target, model
Erroneous detection is less can accurately to filter out inhuman, vehicle target.Monitor video passes through the analysis of this algorithm, realizes to the target in scene
Intelligent recognition, classification and screening, can be significantly reduced the manpower and time cost of manual analysis, while have higher detection again
Accuracy rate, to realize that the intellectual analysis of monitor video is laid a good foundation.
Detailed description of the invention
The attached drawing for constituting a part of the invention is used to provide further understanding of the present invention, schematic reality of the invention
It applies example and its explanation is used to explain the present invention, do not constitute improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is the flow diagram of people's vehicle separation method based on RFCN described in the embodiment of the present invention.
Specific embodiment
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase
Mutually combination.
In the description of the present invention, it is to be understood that, term " center ", " longitudinal direction ", " transverse direction ", "upper", "lower",
The orientation or positional relationship of the instructions such as "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outside" is
It is based on the orientation or positional relationship shown in the drawings, is merely for convenience of description of the present invention and simplification of the description, rather than instruction or dark
Show that signified device or element must have a particular orientation, be constructed and operated in a specific orientation, therefore should not be understood as pair
Limitation of the invention.In addition, term " first ", " second " etc. are used for description purposes only, it is not understood to indicate or imply phase
To importance or implicitly indicate the quantity of indicated technical characteristic.The feature for defining " first ", " second " etc. as a result, can
To explicitly or implicitly include one or more of the features.In the description of the present invention, unless otherwise indicated, " multiple "
It is meant that two or more.
In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, term " installation ", " phase
Even ", " connection " shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or be integrally connected；It can
To be mechanical connection, it is also possible to be electrically connected；It can be directly connected, can also can be indirectly connected through an intermediary
Connection inside two elements.For the ordinary skill in the art, above-mentioned term can be understood by concrete condition
Concrete meaning in the present invention.
The present invention will be described in detail below with reference to the accompanying drawings and embodiments.
Explanation of nouns
RFCN is a kind of network structure.
Anchors indicates set.
Loss value loss function.
A kind of people's vehicle separation method based on RFCN, as shown in Figure 1, including the following steps:
1) video frame images that monitored picture is obtained directly from camera, simply pre-process image.
2) whole image is detected using preparatory trained deep learning detection model, to all in image
People, vehicle target identify and position, and export boundary rectangle, classification and the confidence level of target.
3) 2) testing result in is post-processed, the removal boundary rectangle frame that confidence level is low, degree of overlapping is high, as
Final result output.
4) the continuous video frame images of comprehensive analysis judge the motion state of target, into the time of scene, as judgement
Trigger the reference conditions of warning.
The specific method of step 1) is handled image noise reduction including the use of Mean Filtering Algorithm, if more fuzzy at night
Picture can suitably do acutance processing, then by whole image inspection.
The method of the step 2) detection includes:
201) people, the vehicle sample set of training pattern are constructed；
202) the deep learning network architecture based on RFCN is built；
203) related training parameter, training pattern are configured；
204) detection and the output test result to people, vehicle are realized with training pattern.
Further, step 201) the building people of training pattern, vehicle training sample set specific method include: from
It is obtained in a large amount of actual monitored scene in a variety of actual scenes such as video frame images, including different angle, different light conditions
People, vehicle image with triggering warning alert if.Data enhancing, including rotation, mirror image, sanction are carried out at random to original image
Cut, histogram equalization, gray proces, change brightness, saturation degree, the different enhancement methods such as contrast random combine, thus
Generate sample set abundant.Then marked in the picture using annotation tool people, vehicle boundary rectangle frame and set generic,
Image and corresponding mark file are saved and are used as sample set.
Further, described build based on the RFCN method for constructing the deep learning network architecture of step 202) includes: to adopt
It uses ResNet-18 network as the basic network of image characteristics extraction, the characteristic pattern that basic network exports is input to RFCN mesh
Mark detection framework, the final detection and identification realized to target.Wherein the size and ratio of anchors should meet sample set as far as possible
The scale and proportional region of middle target, convenient for network using prediction block position (x, y, w, h) and confidence level confidence and
The position of true value and classification carry out costing bio disturbance, calculate the loss reduction so that final by iterating.
Further, it is 0.001 that step 203) the configuration training parameter, which is mainly the initial learning rate of setting network,
And every to carry out total four times of training samples number of the number of iterations, learning rate is reduced to original 0.5 times, and total the number of iterations is generally set
It is set to ten times of total number of samples amount.Model training is iterated instruction according to backpropagation thought, using stochastic gradient descent algorithm
Practice, the loss value of network is made to be down to lower value.The loss of network points are Classification Loss L_conf (x, c) and recurrence loss L_loc
(x, l, g), total loss are the weighted sums of the two.Utilize the position l (x, y, w, h) and the value of the confidence of the prediction block of network output
Confidence is calculated with true value g (x, y, w, h) and is lost, and obtains final loss, calculation formula are as follows:
Wherein i indicates i-th of prediction block, and j indicates that target generic, y_ij indicate class represented by i-th of prediction block
Whether do not matched with j-th of classification, otherwise matching as 1 is that 0, x_ij indicates that i-th of prediction block belongs to representated by j-th of true frame
Classification probability, the loss of L_conf (x, c) presentation class.X_i is indicated if between i-th of prediction block and true frame
IOU is greater than 0.7 and is equal to 1, and 0 is equal to when less than 0.3, is otherwise not involved in training.In the actual process, if N_cls and N_loc
The gap of setting is excessive, and can balancing the two with parameter beta, (such as N_cls=256, N_loc=2400, then 10) β may be configured as.
Further, simultaneously output test result is main for the step 204) detection with training pattern realization to people, vehicle
Image to be detected is detected including the use of preparatory trained model, the testing result of model output includes the external of target
Classification and corresponding confidence level belonging to the coordinate of rectangle frame, target.
Post-processing approach described in step 3) mainly includes that a threshold value 0.7 is arranged, goes according to the value of the confidence of each target
Except the lower target of the value of the confidence, caused by these targets may be network erroneous detection.One IOU parameter 0.6 is set, degree of overlapping is removed
The value of the confidence is lower in higher target frame, and what these frames may detect is the same target, only retains a value of the confidence highest
Detection block.According to the size of detection block, removal does not meet the testing result of target sizes range in actual scene.
Step 4) the comprehensive analysis continuous video frame images, mainly according to being exported in the testing result of successive frame
The position of same target judges that the state of the target is static or movement, and can track to the motion profile of target.
According to detecting that the video frame number of same target judges that target enters the time of scene, the reference item as judgement triggering warning
Part.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (9)
1. a kind of people's vehicle separation method based on RFCN, it is characterised in that: specifically comprise the following steps:
1) computer obtains the video frame images of monitored picture directly from camera, is simply pre-processed to image；
2) whole image is detected using preparatory trained deep learning detection model, to all people, vehicle in image
Target is identified and positioned, and boundary rectangle, classification and the confidence level of target are exported；
3) computer post-processes the testing result in 2), and the removal boundary rectangle frame that confidence level is low, degree of overlapping is high is made
It is exported for final result；
4) computer generalization analyzes continuous video frame images, judges the motion state of target, into the time of scene, as sentencing
The reference conditions of disconnected triggering warning.
2. a kind of people's vehicle separation method based on RFCN according to claim 1, it is characterised in that: the tool of the step 1)
Body method is handled image noise reduction including the use of Mean Filtering Algorithm, if more fuzzy picture is suitably done at acutance at night
Reason, then by whole image inspection.
3. a kind of people's vehicle separation method based on RFCN according to claim 2, it is characterised in that: right in the step 2)
The method that whole image is detected includes:
201) people, the vehicle sample set of training pattern are constructed；
202) the deep learning network architecture based on RFCN is built；
203) related training parameter, training pattern are configured；
204) detection and the output test result to people, vehicle are realized with training pattern.
4. a kind of people's vehicle separation method based on RFCN according to claim 3, it is characterised in that: in the step 201)
The specific method of the people, vehicle training sample set that construct training pattern include: to obtain video frame from a large amount of actual monitored scene
Image, including having people, the Che Tu of triggering warning alert if in a variety of actual scenes such as different angle, different light conditions
Picture.Data enhancing carried out at random to original image, including rotation, mirror image, cutting, histogram equalization, gray proces, change it is bright
The random combine of the difference enhancement method such as degree, saturation degree, contrast, to generate sample set abundant.Then using mark work
Tool mark in the picture people, vehicle boundary rectangle frame and set generic, by image and corresponding mark file preservation conduct
Sample set.
5. a kind of people's vehicle separation method based on RFCN according to claim 3, it is characterised in that: in the step 202)
Building the method based on the RFCN building deep learning network architecture includes: using ResNet-18 network as image characteristics extraction
Basic network, the characteristic pattern that basic network exports is input to RFCN target detection frame, the final detection realized to target
And identification；Wherein the size and ratio of anchors should meet the scale and proportional region of target in sample set as far as possible, be convenient for network
Costing bio disturbance is carried out using the position (x, y, w, h) of prediction block, confidence level confidence, the position of true value and classification, is led to
It crosses to iterate and calculates the loss reduction so that final.
6. a kind of people's vehicle separation method based on RFCN according to claim 3, it is characterised in that: in the step 203)
Configuring the initial learning rate that training parameter is mainly setting network is 0.001, and total four times of training samples number of every progress
The number of iterations, learning rate are reduced to original 0.5 times, and total the number of iterations is traditionally arranged to be ten times of total number of samples amount.Model training
According to backpropagation thought, it is iterated training using stochastic gradient descent algorithm, the loss value of network is made to be down to lower value.
7. a kind of people's vehicle separation method based on RFCN according to claim 3, it is characterised in that: in the step 204)
The detection to people, vehicle is realized with training pattern and output test result is mainly including the use of preparatory trained model to be detected
Image is detected, and the testing result of model output includes the coordinate of boundary rectangle frame of target, classification belonging to target and right
The confidence level answered.
8. a kind of people's vehicle separation method based on RFCN according to claim 3, it is characterised in that: right in the step 3)
The method that testing result is post-processed mainly includes that a threshold value is arranged according to the value of the confidence of each target, removes the value of the confidence
An IOU parameter is arranged in lower target, and the value of the confidence is lower in the removal higher target frame of degree of overlapping, these frames may be examined
What is surveyed is the same target, only retains a highest detection block of the value of the confidence.According to the size of detection block, removal is not met
The testing result of target sizes range in actual scene.
9. a kind of people's vehicle separation method based on RFCN according to claim 1, it is characterised in that: comprehensive in the step 4)
It closes and analyzes continuous video frame images, mainly according to the position of the same target exported in the testing result of successive frame, judgement
The state of the target is static or movement, and can be tracked to the motion profile of target.According to detecting same target
Video frame number judge that target enters the time of scene, as judgement triggering warning reference conditions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811585378.7A CN109711322A (en) | 2018-12-24 | 2018-12-24 | A kind of people's vehicle separation method based on RFCN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811585378.7A CN109711322A (en) | 2018-12-24 | 2018-12-24 | A kind of people's vehicle separation method based on RFCN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109711322A true CN109711322A (en) | 2019-05-03 |
Family
ID=66257293
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811585378.7A Pending CN109711322A (en) | 2018-12-24 | 2018-12-24 | A kind of people's vehicle separation method based on RFCN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109711322A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110519485A (en) * | 2019-09-09 | 2019-11-29 | Oppo广东移动通信有限公司 | Image processing method, device, storage medium and electronic equipment |
| CN110633680A (en) * | 2019-09-19 | 2019-12-31 | 天津天地伟业机器人技术有限公司 | Video-based people number abnormity detection method |
| CN110688924A (en) * | 2019-09-19 | 2020-01-14 | 天津天地伟业机器人技术有限公司 | RFCN-based vertical monocular passenger flow volume statistical method |
| CN110866512A (en) * | 2019-11-21 | 2020-03-06 | 南京大学 | Monitoring camera shielding detection method based on video classification |
| CN111046797A (en) * | 2019-12-12 | 2020-04-21 | 天地伟业技术有限公司 | Oil pipeline warning method based on personnel and vehicle behavior analysis |
| CN111046822A (en) * | 2019-12-19 | 2020-04-21 | 山东财经大学 | Large vehicle anti-theft method based on artificial intelligence video identification |
| CN111950517A (en) * | 2020-08-26 | 2020-11-17 | 司马大大（北京）智能***有限公司 | Target detection method, model training method, electronic device and storage medium |
| CN112183558A (en) * | 2020-09-30 | 2021-01-05 | 北京理工大学 | Target detection and feature extraction integrated network based on YOLOv3 |
| CN112465691A (en) * | 2020-11-25 | 2021-03-09 | 北京旷视科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable medium |
| CN113554008A (en) * | 2021-09-18 | 2021-10-26 | 深圳市安软慧视科技有限公司 | Method and device for detecting static object in area, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062349A (en) * | 2017-10-31 | 2018-05-22 | 深圳大学 | Video frequency monitoring method and system based on video structural data and deep learning |
| CN108197613A (en) * | 2018-02-12 | 2018-06-22 | 天津天地伟业信息***集成有限公司 | A kind of Face datection optimization algorithm based on depth convolution cascade network |
| CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
| CN108427920A (en) * | 2018-02-26 | 2018-08-21 | 杭州电子科技大学 | A kind of land and sea border defense object detection method based on deep learning |
| CN108875595A (en) * | 2018-05-29 | 2018-11-23 | 重庆大学 | A kind of Driving Scene object detection method merged based on deep learning and multilayer feature |
-
2018
- 2018-12-24 CN CN201811585378.7A patent/CN109711322A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062349A (en) * | 2017-10-31 | 2018-05-22 | 深圳大学 | Video frequency monitoring method and system based on video structural data and deep learning |
| CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
| CN108197613A (en) * | 2018-02-12 | 2018-06-22 | 天津天地伟业信息***集成有限公司 | A kind of Face datection optimization algorithm based on depth convolution cascade network |
| CN108427920A (en) * | 2018-02-26 | 2018-08-21 | 杭州电子科技大学 | A kind of land and sea border defense object detection method based on deep learning |
| CN108875595A (en) * | 2018-05-29 | 2018-11-23 | 重庆大学 | A kind of Driving Scene object detection method merged based on deep learning and multilayer feature |
Non-Patent Citations (2)
| Title |
|---|
| FENG ZHANG 等: "Detection of Road Surface Identifiers Based on Deep Learning", 《2018 3RD INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND INFORMATION ENGINEERING》 * |
| 李超凡等: "基于学习算法SSD的实时道路拥堵检测", 《软件导刊》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110519485A (en) * | 2019-09-09 | 2019-11-29 | Oppo广东移动通信有限公司 | Image processing method, device, storage medium and electronic equipment |
| CN110519485B (en) * | 2019-09-09 | 2021-08-31 | Oppo广东移动通信有限公司 | Image processing method, image processing device, storage medium and electronic equipment |
| CN110633680A (en) * | 2019-09-19 | 2019-12-31 | 天津天地伟业机器人技术有限公司 | Video-based people number abnormity detection method |
| CN110688924A (en) * | 2019-09-19 | 2020-01-14 | 天津天地伟业机器人技术有限公司 | RFCN-based vertical monocular passenger flow volume statistical method |
| CN110866512A (en) * | 2019-11-21 | 2020-03-06 | 南京大学 | Monitoring camera shielding detection method based on video classification |
| CN111046797A (en) * | 2019-12-12 | 2020-04-21 | 天地伟业技术有限公司 | Oil pipeline warning method based on personnel and vehicle behavior analysis |
| CN111046822A (en) * | 2019-12-19 | 2020-04-21 | 山东财经大学 | Large vehicle anti-theft method based on artificial intelligence video identification |
| CN111950517A (en) * | 2020-08-26 | 2020-11-17 | 司马大大（北京）智能***有限公司 | Target detection method, model training method, electronic device and storage medium |
| CN112183558A (en) * | 2020-09-30 | 2021-01-05 | 北京理工大学 | Target detection and feature extraction integrated network based on YOLOv3 |
| CN112465691A (en) * | 2020-11-25 | 2021-03-09 | 北京旷视科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable medium |
| CN113554008A (en) * | 2021-09-18 | 2021-10-26 | 深圳市安软慧视科技有限公司 | Method and device for detecting static object in area, electronic equipment and storage medium |
| CN113554008B (en) * | 2021-09-18 | 2021-12-31 | 深圳市安软慧视科技有限公司 | Method and device for detecting static object in area, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109711322A (en) | A kind of people's vehicle separation method based on RFCN | |
| CN110135269B (en) | Fire image detection method based on mixed color model and neural network | |
| CN113255481B (en) | Crowd state detection method based on unmanned patrol car | |
| CN104378582B (en) | A kind of intelligent video analysis system and method cruised based on Pan/Tilt/Zoom camera | |
| CN105678811B (en) | A kind of human body anomaly detection method based on motion detection | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| WO2017190574A1 (en) | Fast pedestrian detection method based on aggregation channel features | |
| CN100545867C (en) | Aerial shooting traffic video frequency vehicle rapid checking method | |
| CN111325713A (en) | Wood defect detection method, system and storage medium based on neural network | |
| CN109670441A (en) | A kind of realization safety cap wearing knows method for distinguishing, system, terminal and computer readable storage medium | |
| CN108596128B (en) | Object recognition method, device and storage medium | |
| CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
| CN108537215A (en) | A kind of flame detecting method based on image object detection | |
| CN105243356B (en) | A kind of method and device that establishing pedestrian detection model and pedestrian detection method | |
| CN108932479A (en) | A kind of human body anomaly detection method | |
| CN104320617B (en) | A kind of round-the-clock video frequency monitoring method based on deep learning | |
| CN104463869B (en) | A kind of video flame image composite identification method | |
| CN105893946A (en) | Front face image detection method | |
| CN106778540B (en) | Parking detection is accurately based on the parking event detecting method of background double layer | |
| CN110255318B (en) | Method for detecting idle articles in elevator car based on image semantic segmentation | |
| CN109815863A (en) | Firework detecting method and system based on deep learning and image recognition | |
| CN112183472A (en) | Method for detecting whether test field personnel wear work clothes or not based on improved RetinaNet | |
| CN113192038B (en) | Method for recognizing and monitoring abnormal smoke and fire in existing flame environment based on deep learning | |
| Malhi et al. | Vision based intelligent traffic management system | |
| Tao et al. | Smoky vehicle detection based on range filtering on three orthogonal planes and motion orientation histogram |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190503 |
|
| RJ01 | Rejection of invention patent application after publication |