CN113793508B - Anti-interference rapid detection method for entrance and exit unlicensed vehicle - Google Patents
Anti-interference rapid detection method for entrance and exit unlicensed vehicle Download PDFInfo
- Publication number
- CN113793508B CN113793508B CN202111134956.7A CN202111134956A CN113793508B CN 113793508 B CN113793508 B CN 113793508B CN 202111134956 A CN202111134956 A CN 202111134956A CN 113793508 B CN113793508 B CN 113793508B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- gradient
- background
- image
- detection area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
- G08G1/0175—Detecting movement of traffic to be counted or controlled identifying vehicles by photographing vehicles, e.g. when violating traffic rules
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an anti-interference rapid detection method for a license-free vehicle at an entrance, which is applied to capturing of the license-free vehicle at a charging scene at the entrance, and comprises the following steps: firstly, establishing a gradient background of a detection area; then, detecting the running state of the vehicle, synchronously acquiring a background gradient map, a current frame and a background gradient difference map, and judging the state of the vehicle through the image change condition; in the no-vehicle state, the background is continuously updated. The patent discloses a rapid and anti-interference vehicle detection method based on combination of various methods such as vehicle characteristic points, optical flow, a frame difference method, gradient background reconstruction and the like. It can improve vehicle access speed and accuracy.
Description
Technical Field
The invention relates to a vehicle detection method, in particular to an anti-interference rapid detection method for a license-free vehicle at an entrance.
Background
In the parking charging field of the gateway, the charging system generally comprises the following steps: when a vehicle is driven into a parking lot, capturing vehicle information, and recording the entrance time; when the vehicle exits the parking lot, inquiring the vehicle entrance time, and calculating the payment amount. Usually, the vehicle is driven into and recorded according to license plate number recognition, but vehicles without license plates are frequently encountered, and because of the occurrence of the license plate-free vehicles, automatic snapshot and charging of the vehicles are not feasible only by virtue of the license plate recognition, auxiliary snapshot is required to be carried out by virtue of vehicle detection, target detection is carried out by virtue of the vehicle detection, and then the structural information (vehicle color, vehicle brand, vehicle system and vehicle type) of the vehicles is recognized, and the structural information is recorded as the description information of the license-free vehicles; when the vehicle exits, inquiring the description information of the entering vehicle recorded in the database, matching the vehicle with the highest similarity with the vehicle, and carrying out charging calculation.
In the flow, when license plate snapshot fails, a vehicle snapshot mechanism is started. There are many factors that interfere with the detection of vehicles in practical situations, such as pedestrians, bicycles, motorcycles, abrupt light changes, etc. In addition, the programs are all operated in the camera, and the calculation resources are limited, so that a rapid anti-interference vehicle detection algorithm must be developed to well solve the snapshot charging problem of the unlicensed vehicle.
Disclosure of Invention
The invention aims to solve the main technical problem of providing an anti-interference rapid detection method for a license-free vehicle at an entrance and an exit, which can improve the detection speed and accuracy of the vehicle.
In order to solve the technical problems, the invention provides an anti-interference rapid detection method for a non-license vehicle at an entrance, which comprises the following steps:
when the license plate snapshot fails, a vehicle snapshot mechanism is started.
Step one, establishing a gradient background of a detection area: based on the video image, defining a vehicle detection area in a vehicle-free state, establishing a gradient background of the detection area, and continuously updating the background in the vehicle-free state;
step two, detecting the running state of the vehicle: synchronously acquiring a background gradient map, calculating the gradient of a detection region of a current frame and a background gradient difference (a current frame and a background gradient difference map), and calculating the gradient difference of the detection region of the current frame and a previous frame (a gradient difference map of a previous frame and a previous frame);
if the current frame, the background gradient difference image and the front and rear frame gradient difference images are obviously changed, the vehicle is changed from a vehicle-free state to a vehicle-mounted state, and the vehicle is judged to be driven in;
if the gradient difference images of the current frame and the background have obvious changes, judging that the vehicle stops when the gradient difference images of the front frame and the rear frame have no changes;
if the current frame, the background gradient difference image and the front and rear frame gradient difference images are obviously changed and are in a vehicle state all the time, judging that the vehicle runs;
if the gradient difference images of the current frame and the background and the gradient difference images of the front frame and the rear frame have no obvious change, judging that the vehicle has driven out.
And step three, continuously updating the background in the vehicle-free state. (in the absence of a car, every frame may be in real-time update background.)
When the vehicle is judged to be stopped, the vehicle structural information is identified, and the vehicle color, the vehicle brand, the vehicle system and the vehicle type are obtained and used as the vehicle description information.
Comparing the description information of the vehicle before the vehicle description information is stored with the description information of the stored vehicle; when the same situation does not occur, directly storing; when the same situation occurs, the driver-identifying image information is added as additional information of the vehicle description information, and then stored.
The driver profile information includes at least one of the following factors: long and short hair, facial aspect ratio, eyebrow spacing, and nose-mouth spacing.
When the vehicle information is the same again, a driver image information factor is increased. That is, only one driver image information factor is taken, two driver image information factors are taken now, and the vehicles are distinguished.
In the implementation, in order to ensure the real-time performance of the background, the information (1/50 information weight) of the current image is gradually increased on the basis of the original background A, so that the background light is gradually updated along with the change of time.
When in implementation, the method also comprises an anti-interference processing flow before the second step, which comprises the following steps:
a gradient image is adopted to establish a background, and a gradient calculation method adopts a gradient in the vertical direction; (treatment for preventing abrupt change of light)
Dividing the detection area into an upper part, a middle part, a lower part and a third part, and judging that the vehicle is in a top-down or top-down entering state;
the object area is too small, and judging that the vehicle is not a vehicle;
and comparing whether the calculated direction of the motion vector is consistent with the preset direction of the vehicle to judge whether the vehicle is the vehicle.
In implementation, the method further comprises a rapid operation step after the second step:
reducing data processing, wherein an algorithm directly reduces the acquired image of the camera to 200 x 110 pixels, and simultaneously processes a detection area part in the image; the detection area accounts for about 1/3 of the image;
the rapid operation step further comprises rapid detection of characteristic points, and rapid detection of special detection points is achieved through table lookup.
The beneficial effects of the invention are as follows: the method for rapidly detecting the anti-interference of the entrance and exit license-free vehicles comprises the following steps: firstly, establishing a gradient background of a detection area; then, detecting the running state of the vehicle, synchronously acquiring a background gradient map, a current frame and a background gradient difference map, and judging the state of the vehicle through the image change condition; in the no-vehicle state, the background is continuously updated. The patent discloses a rapid and anti-interference vehicle detection method based on combination of various methods such as vehicle characteristic points, optical flow, a frame difference method, gradient background reconstruction and the like. It can improve vehicle access speed and accuracy.
Drawings
In order to more clearly illustrate the technical solutions of the present application, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a gradient background establishment of a detection region according to an embodiment of the present invention;
FIG. 2a is a schematic diagram of a background gradient of an embodiment of the present invention;
FIG. 2b is a schematic diagram of the current frame and background gradient difference in accordance with an embodiment of the present invention;
FIG. 2c is a schematic diagram of the gradient difference between the front and rear frames according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a vehicle running state detection flow according to an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
Referring to fig. 1 to 3, the present invention discloses a method for rapidly detecting the anti-interference of a license-free vehicle at an entrance, which is based on a combination of a plurality of methods such as vehicle feature points, optical flow, a frame difference method, and gradient background reconstruction. The method is mainly applied to the non-license vehicle snapshot of the charging scene at the entrance and the exit. The implementation steps are as follows:
1. vehicle inspection
1. And establishing a gradient background of the detection area. As shown in fig. 1, the trapezoid area is a vehicle detection area, a gradient background of the current area is established in a no-vehicle state, and the background is continuously updated in the no-vehicle state.
Gradient background refers to gradient information of gray level images in the current processing area, and the gradient calculation method adopted herein adopts the following calculation method: around the current point there are 8 points, the three points below are subtracted from the three points above, the absolute values are summed, and the average is calculated as follows:
a＝(abs(a1-a4)+abs(a2-a5)+abs(a3-a6))/3，
where a1, a2, a3 are three points above the current point and a4, a5, a6 are three points below the current point.
Because the vehicle head has a plurality of transverse textures, more information can be obtained through subtraction in the vertical direction, the operation can be reduced by omitting horizontal subtraction, and meanwhile, the interference of objects with strong longitudinal textures such as pedestrians can be reduced.
| a1 | a2 | a3 |
| a | ||
| a4 | a5 | a6 |
(the above table is for the purpose of vivid presentation of the pixel point positions of the above formula a= (abs (a 1-a 4) +abs (a 2-a 5) +abs (a 3-a 6))/3, a is the current position, a1, a2, a3 are three points above the current point, a4, a5, a6 are three points below the current point)
And updating the gradient background, calculating a gradient image of each frame of image detection area in a vehicle-free state, and adding weights with the current gradient background. The formula is as follows:
A＝0.98*A+0.02*B，
a is the current gradient background, B is the gradient map of the current frame detection area, and the corresponding pixels of the two images are updated by weighting through the above formula.
The purpose of the combination of the steps is to ensure the real-time performance of the background, gradually increase the information (1/50 information weight) of the current image on the basis of the original background A, and ensure that the background light is gradually updated along with the change of time.
2. Vehicle running state detection: vehicle entrance, vehicle stop, vehicle travel, vehicle exit four states (here entrance and exit refers to a vehicle entrance camera scene and an exit camera scene). The method comprises the following specific steps:
1) Calculating the gradient of the detection area of the current frame and the gradient difference of the background;
2) Calculating the gradient difference of the detection areas of the current frame and the previous frame; as shown in fig. 2a, 2b, 2c, background gradient-fig. 2a, middle is current frame and background gradient difference-fig. 2b, lower is front-to-back frame gradient difference-fig. 2c.
3) Judging that the vehicle is driven in, and when the vehicle is driven in, the figure 2b and the figure 2c are obviously changed;
4) Judging that the vehicle stops, when the vehicle stops, the graph 2b has obvious change, and the graph 2c has no change;
5) The vehicle running judgment has the same effect as 3), and the difference is that the vehicle is in a vehicle-free state to a vehicle-in state, and the vehicle is in a vehicle-in state all the time when the vehicle is in the vehicle-in state;
6) Judging the vehicle driving out, wherein after the vehicle is driven out, both the figure 2b and the figure 2c have no change;
7) In the no-vehicle state, the background is continuously updated.
2. The anti-interference treatment (compared with the prior art, the change is that 1, the gray background is changed into the gradient background to prevent the light mutation 2, the original detection area is divided into an upper part, a middle part and a lower part, and meanwhile, the light flow is added to accurately judge the movement direction) mainly interferes with the light mutation, pedestrians, bicycles and motorcycles.
1. Light ray mutation treatment
1) Because the gray level image is greatly affected by abrupt change of light, a gradient image is used to build up a background, the light suddenly lightens or darkens, and the gradient background remains stable.
2) The gradient calculation method comprises the following steps: because of pedestrians or motorcycles, the gradient in the horizontal direction is large (vertical) and the gradient in the vertical direction of the vehicle is large (the head portion is more lateral), in order to reduce the disturbance, the gradient in the vertical direction is employed.
2. Treatment of moving objects such as pedestrians, bicycles, motorcycles, and the like
1) The pedestrians, bicycles and motorcycles can move more randomly, and the automobiles can be fixed more. Therefore, the detection area is divided into three parts, namely an upper part, a middle part and a lower part, the vehicle is driven in from top to bottom or from bottom to top, other objects are irregular, and partial random moving objects can be eliminated according to the change sequence of the three parts.
2) And filtering the object area. Since the disturbance is generally small relative to the vehicle, the pass-through area can filter a large portion of the disturbance.
3) Accurate filtering of motion direction
The above 1) and 2) can filter most of interference, but cannot completely filter, so that feature point detection and optical flow are added to make accurate judgment.
And (3) detecting the characteristic points rapidly, wherein the light color points in the image are characteristic point effects on the vehicle.
And judging the running direction of the feature points by using the optical flow, and obtaining a motion vector by linking two points if the deep color point in the figure is the position of a frame on the current shallow color point.
And counting the motion vectors by using the histogram, wherein the position where the motion vectors are gathered is the accurate direction of the motion of the vehicle. Since the direction of movement of the vehicle can be preset: upward, downward (finer can be set to left, right, up) and comparing whether the calculated direction of the motion vector is consistent with the preset direction of the vehicle to judge whether the vehicle is the vehicle.
3. Fast operation
As the front-end equipment performance is general, high-performance operations such as AI and the like are not supported, and the algorithm speed is very strict.
1) The data processing is reduced, the camera acquires images 1920 x 1080 pixels, the algorithm directly reduces the images to 200 x 110 pixels (as the motion change of the whole vehicle is processed and details are not processed, the whole outline is still obvious after the images are reduced, the accuracy rate is not reduced), and meanwhile, the detection area in the images is partially processed. The detection area occupies about 1/3 of the image, so the processing data volume is only 200×110/3 pixels, and the information such as gradient can be calculated on the data volume very quickly.
2) And (3) rapidly detecting the characteristic points, and realizing rapid special detection point by looking up a table.
Please refer to the detection flow: 2-step vehicle running state detection in vehicle detection, as shown in the flowchart 2, is as follows:
the camera of the gateway inputs a color image,
when the vehicle is judged to be stopped, the vehicle structural information is identified, and the vehicle color, the vehicle brand, the vehicle system and the vehicle type are obtained and used as the vehicle description information. If two identical unlicensed vehicles exist, but the probability that the colors, brands, trains and models of the identical unlicensed vehicles are identical is very small, so that the unlicensed vehicles can be distinguished through the information.
Comparing the description information of the vehicle before the vehicle description information is stored with the description information of the stored vehicle; when the same situation does not occur, directly storing; when the same situation occurs, the driver-identifying image information is added as additional information of the vehicle description information, and then stored. When two identical license-free vehicles happen to appear, the license-free vehicles cannot be distinguished basically through the information of the vehicles, and meanwhile, distinguishable additional information can be added in percentage effectively through adding the information of the identified drivers.
The driver profile information includes the following factors: long and short hair, facial aspect ratio, eyebrow spacing, and nose-mouth spacing. The driver's image information takes at least one factor (may be cis-position recognition) in front (when the same vehicle description information still appears and the same driver's image information, one more image information factor is added to recognize the at least two, and the same more image information factor is added to recognize the at least three more image information factors). Such information has low requirements on image sharpness and is effective in distinguishing most people. Therefore, under the condition of combining at least two image information, vehicles can be effectively distinguished, the recognition speed is high, the image requirement is low, and the difference of different individual related information is large, so that the recognition degree is improved. Even if the same driver image information factors of different vehicles cannot be completely distinguished, a certain distinguishing effect can be realized by changing the quantity of the driver image information factors alone.
When a non-licensed vehicle enters the garage, the vehicle description information is stored to the warehoused vehicle database along with the additional information. When the non-license vehicle exits the garage, the identification related information is compared with the data in the database of the warehoused vehicles, and the warehouse-in time of the corresponding vehicle is found for charging.
In the schematic diagram of the vehicle running state detection flow shown in fig. 3, thr is an empirical value summarized according to an actual test, and different empirical values are counted according to different scenes.
The foregoing is a further detailed description of the invention in connection with specific embodiments, and it is not intended that the invention be limited to such description. It will be apparent to those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the invention, and these should be considered to be within the scope of the invention.
Claims (4)
1. A method for rapidly detecting the anti-interference of a license plate-free vehicle at an entrance is characterized in that a vehicle snapshot mechanism is started when license plate snapshot fails, and the specific method comprises the following steps:
step one, establishing a gradient background of a detection area: based on the video image, defining a vehicle detection area in a vehicle-free state, establishing a gradient background of the detection area, and continuously updating the background in the vehicle-free state;
the anti-interference processing flow is included before the second step, and specifically includes: a gradient image is adopted to establish a background, and a gradient calculation method adopts a gradient in the vertical direction; the anti-interference processing flow further comprises: dividing the detection area into an upper part, a middle part, a lower part and a third part, and judging that the vehicle is in a top-down or top-down entering state; comparing whether the calculated direction of the motion vector is consistent with the preset direction of the vehicle to judge whether the vehicle is the vehicle;
step two, detecting the running state of the vehicle: synchronously acquiring a background gradient map, calculating the gradient of a detection area of a current frame and a background gradient difference, and calculating the gradient difference of the detection area of the current frame and a detection area of a previous frame;
if the current frame, the background gradient difference image and the front and rear frame gradient difference images are obviously changed, the vehicle is changed from a vehicle-free state to a vehicle-mounted state, and the vehicle is judged to be driven in;
if the gradient difference images of the current frame and the background have obvious changes, judging that the vehicle stops when the gradient difference images of the front frame and the rear frame have no changes;
if the current frame, the background gradient difference image and the front and rear frame gradient difference images are obviously changed and are in a vehicle state all the time, judging that the vehicle runs;
if the gradient difference images of the current frame and the background and the gradient difference images of the front frame and the rear frame have no obvious change, judging that the vehicle is driven out;
the method comprises the following steps of fast operation: reducing data processing, wherein an algorithm directly reduces the acquired image of the camera to 200 x 110 pixels, and simultaneously processes a detection area part in the image; the detection area accounts for about 1/3 of the image; the method also comprises the steps of rapidly detecting the characteristic points, and realizing rapid special detection point detection through table lookup;
when the vehicle is judged to be stopped, vehicle structural information identification is carried out, and vehicle colors, vehicle brands, vehicle systems and vehicle types are obtained and used as vehicle description information; comparing the description information of the vehicle before the vehicle description information is stored with the description information of the stored vehicle; when the same situation does not occur, directly storing; when the same situation occurs, the image information of the driver is additionally identified and used as the additional information of the vehicle description information, and then the additional information is stored;
and step three, continuously updating the background in the vehicle-free state.
2. The method for rapidly detecting the anti-interference of the license-free vehicles at the gateway of claim 1, wherein the anti-interference processing flow further comprises: and judging that the vehicle is not a vehicle when the object area is too small.
3. The entrance-and-exit license-free vehicle tamper-proof rapid detection method of claim 1, wherein the driver image information factor comprises: at least one of long and short hair, facial aspect ratio, eyebrow spacing, and nose-mouth spacing.
4. The tamper-proof rapid detection method for entrance/exit license-free vehicles according to claim 3, wherein when the vehicle information is the same again, a driver's image information factor is added.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111134956.7A CN113793508B (en) | 2021-09-27 | 2021-09-27 | Anti-interference rapid detection method for entrance and exit unlicensed vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111134956.7A CN113793508B (en) | 2021-09-27 | 2021-09-27 | Anti-interference rapid detection method for entrance and exit unlicensed vehicle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113793508A CN113793508A (en) | 2021-12-14 |
| CN113793508B true CN113793508B (en) | 2023-06-16 |
Family
ID=79184501
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111134956.7A Active CN113793508B (en) | 2021-09-27 | 2021-09-27 | Anti-interference rapid detection method for entrance and exit unlicensed vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113793508B (en) |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101872546B (en) * | 2010-05-06 | 2013-03-13 | 复旦大学 | Video-based method for rapidly detecting transit vehicles |
| CN102496276B (en) * | 2011-12-01 | 2013-08-21 | 青岛海信网络科技股份有限公司 | High efficiency vehicle detection method |
| CN103337175A (en) * | 2013-06-22 | 2013-10-02 | 太仓博天网络科技有限公司 | Vehicle type recognition system based on real-time video steam |
| KR101647370B1 (en) * | 2014-11-26 | 2016-08-10 | 휴앤에스(주) | road traffic information management system for g using camera and radar |
| CN106205139B (en) * | 2016-07-14 | 2019-03-08 | 浙江宇视科技有限公司 | Vehicle room entry/exit management method and device |
| CN106448184B (en) * | 2016-12-15 | 2019-03-01 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification method and vehicle appearance recognition methods |
| CN109034171B (en) * | 2018-07-24 | 2022-10-04 | 北京易泊时代科技有限公司 | Method and device for detecting unlicensed vehicles in video stream |
| CN109598246B (en) * | 2018-12-07 | 2023-01-31 | 广东亿迅科技有限公司 | Vehicle access detection method and device, computer equipment and storage medium |
| CN109448147A (en) * | 2018-12-12 | 2019-03-08 | 杭州华云科技有限公司 | A kind of unlicensed vehicle parking charging method and parking charge system |
| CN109684996B (en) * | 2018-12-22 | 2020-12-04 | 北京工业大学 | Real-time vehicle access identification method based on video |
| CN210777138U (en) * | 2019-11-07 | 2020-06-16 | 深圳前海华夏智信数据科技有限公司 | Vehicle passing identification equipment and parking management system |
-
2021
- 2021-09-27 CN CN202111134956.7A patent/CN113793508B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113793508A (en) | 2021-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109829403B (en) | Vehicle anti-collision early warning method and system based on deep learning | |
| CN109784162B (en) | Pedestrian behavior recognition and trajectory tracking method | |
| CN108725440B (en) | Forward collision control method and apparatus, electronic device, program, and medium | |
| CN106652465B (en) | Method and system for identifying abnormal driving behaviors on road | |
| US8744132B2 (en) | Video-based method for detecting parking boundary violations | |
| Takeuchi et al. | On-road vehicle tracking using deformable object model and particle filter with integrated likelihoods | |
| US10878259B2 (en) | Vehicle detecting method, nighttime vehicle detecting method based on dynamic light intensity and system thereof | |
| CN107944388A (en) | A kind of method for detecting lane lines, device and terminal | |
| CN111553214A (en) | Method and system for detecting smoking behavior of driver | |
| CN104951758A (en) | Vehicle-mounted method and vehicle-mounted system for detecting and tracking pedestrians based on vision under urban environment | |
| CN103927548A (en) | Novel vehicle collision avoiding brake behavior detection method | |
| CN112261390A (en) | Vehicle-mounted camera equipment and image optimization device and method thereof | |
| KR20180127245A (en) | Method for Predicting Vehicle Collision Using Data Collected from Video Games | |
| Xiao et al. | Detection of drivers visual attention using smartphone | |
| CN113793508B (en) | Anti-interference rapid detection method for entrance and exit unlicensed vehicle | |
| CN116740792A (en) | Face recognition method and system for sightseeing vehicle operators | |
| Chen et al. | Vehicle classification and counting system | |
| CN113486837B (en) | Automatic driving control method for low-pass obstacle | |
| Vijverberg et al. | High-level traffic-violation detection for embedded traffic analysis | |
| CN114255452A (en) | Target ranging method and device | |
| Gao et al. | An In-Car Objects Detection Algorithm Based on Improved Spatial-Temporal Entropy Image | |
| CN116152790B (en) | Safety belt detection method and device | |
| Qian et al. | Obstacles extraction using a moving camera | |
| Zhan et al. | Vehicle Forward Collision Warning Based on Improved Deep Neural Network | |
| CN117636122A (en) | Training method and detection method for vehicle type axle recognition |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |