CN106373394B - A kind of vehicle checking method and system based on video and radar - Google Patents

A kind of vehicle checking method and system based on video and radar Download PDF

Info

Publication number
CN106373394B
CN106373394B CN201610816421.0A CN201610816421A CN106373394B CN 106373394 B CN106373394 B CN 106373394B CN 201610816421 A CN201610816421 A CN 201610816421A CN 106373394 B CN106373394 B CN 106373394B
Authority
CN
China
Prior art keywords
vehicle
video
based
radar
graphic
Prior art date
Application number
CN201610816421.0A
Other languages
Chinese (zh)
Other versions
CN106373394A (en
Inventor
李纪奎
Original Assignee
深圳尚桥交通技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳尚桥交通技术有限公司 filed Critical 深圳尚桥交通技术有限公司
Priority to CN201610816421.0A priority Critical patent/CN106373394B/en
Publication of CN106373394A publication Critical patent/CN106373394A/en
Application granted granted Critical
Publication of CN106373394B publication Critical patent/CN106373394B/en

Links

Abstract

The present invention relates to a kind of vehicle checking method and system based on video and radar, wherein method is the following steps are included: acquire the radar data of section vehicle to be detected;Acquire the video image in section to be detected;Based on the radar data and video image, the vehicle graphic radar data corresponding with vehicle on section to be detected on video image is matched, and calculates vehicle detection index;The calculated result of vehicle detection index is superimposed upon on video image and is shown.The present invention is by blending video and detections of radar, both the visualization of radar data can have been realized by superposition of data on the video images, it can use radar data again and solve the problems, such as that video detection is affected by environment big, make detection data more comprehensively, the accuracy rate of detection also greatly improves.

Description

A kind of vehicle checking method and system based on video and radar

Technical field

The present invention relates to technical field of vehicle detection more particularly to a kind of vehicle checking method based on video and radar and System.

Background technique

As urban population is more and more intensive, vehicle is more and more, and alleviating urban traffic blocking becomes each urban development Urgent problem.Real-time optimization municipal traffic control signal, the travel time for reducing people are that reach this purpose effective Means, and it is round-the-clock, real-time, comprehensively and accurately information of vehicles detection device is the inevitable approach of this means.

Wagon detector currently on the market mainly has the coil checker of contact, geomagnetism detecting device and contactless Video detector and radar detector.For contact measurement device, maximum defect is to need to destroy road surface to be pacified Dress, detection data are relatively simple.So noncontacting proximity sensor is mainstream wagon detector at present and in the future.

The effect of visualization of video-based vehicle detection is good, can monitor in real time road vehicle, in real time viewing detection effect with And vehicle running track.In addition, it is also relatively more comprehensive based on the Testing index that video data obtains, including vehicle flowrate, occupation rate, row The indexs such as team leader's degree, headstock distance, and do not influenced by vehicle target quantity.Video-based vehicle detection can also detect movement With static all vehicles.But video-based vehicle detection does not adapt to various weather environments, such as dense fog, heavy rain, heavy snow Etc. evils omit weather under, verification and measurement ratio is greatly reduced, and is unable to normal use.And video-based vehicle detection can not accurately detect vehicle Instantaneous velocity.

Radar vehicle detector is not influenced then by weather environment and light, can be truly realized round-the-clock vehicle inspection It surveys, and also can detecte the indexs such as vehicle flowrate, instantaneous radial speed, vehicle radial position information.But radar vehicle detects Device has no idea to visualize, and can't see the image information of vehicle.When vehicle congestion, that is, destination number is excessive When, verification and measurement ratio will decline, and cannot detect static or too low speed vehicle.

Summary of the invention

The technical problem to be solved by the present invention is to detect for existing single video-based vehicle detection or radar vehicle There is respective deficiency in device.A kind of vehicle checking method and system based on video and radar is provided.

In order to solve the above-mentioned technical problems, the present invention provides a kind of vehicle checking method based on video and radar, packet Include following steps:

Acquire the radar data of section vehicle to be detected;

Acquire the video image in section to be detected;

Based on the radar data and video image, by the vehicle graphic on video image and vehicle pair on section to be detected The radar data answered matches, and calculates vehicle detection index;

The calculated result of vehicle detection index is superimposed upon on video image and is shown.

In the vehicle checking method according to the present invention based on video and radar, the vehicle by video image The step of figure radar data corresponding with vehicle on section to be detected matches includes: to be detected based on video image identification It whether there is vehicle graphic in region, be to obtain the vehicle graphic coordinate of vehicle graphic on the video images, the vehicle figure Shape coordinate includes abscissa rx' and ordinate ry', when according to ordinate ry' judge that the vehicle graphic is in the area to be detected When in the start line delimited in advance in domain, according to the abscissa r of vehicle graphicx' and the area to be tested of measured in advance on it is each The lateral distance XA of a lane line1' to XAn' judge the vehicle graphic lane information, if XAi’≤|rx’|<XAi+1', then sentence The vehicle graphic that breaks belongs to i-th of lane, 1≤i < n;Wherein n is the quantity in lane in area to be tested;It is mentioned based on radar data The actual position coordinate of pick-up, the actual position coordinate includes abscissa rxAnd ordinate ry, when according to ordinate rySentence The vehicle break when being in the start line delimited in advance in section to be detected, according to the abscissa r of vehiclexAnd measured in advance The lateral distance XA of each lane line on section to be detected1To XAnThe lane information of the vehicle is judged, if XAj≤|rx|< XAj+1, then judge that the vehicle belongs to j-th of lane, 1≤j < n;Wherein n is the quantity in lane on section to be detected;As i=j, The radar data of the vehicle graphic that will identify that and the vehicle matches.

In the vehicle checking method according to the present invention based on video and radar, the vehicle detection index includes Environment visibility, vehicle holding time, vehicle location, lane information, car speed, queue length, flow and/or road Situation.

In the vehicle checking method according to the present invention based on video and radar, the calculating vehicle detection index The step of include following vehicle holding time calculating step: environment visibility is calculated based on video image;Judge environment visibility Whether it is higher than default visibility, is, whether there is based on video images detection vehicle, and record the holding time;Otherwise it is based on thunder Up to the Data Detection vehicle holding time.

It is described to be based on video images detection in the vehicle checking method according to the present invention based on video and radar It includes: LBP operator extraction vehicle textural characteristics using 9 × 9 neighborhoods that vehicle, which whether there is, after being trained to sample image To the strong classifier for having LBP feature, judged in video image using the strong classifier with LBP feature with the presence or absence of vehicle Whether figure is had the primary testing result of vehicle;When the primary testing result detection has vehicle, HAAR feature extraction is used Vehicle textural characteristics obtain the strong classifier with HAAR feature after being trained to sample image, special using that should have HAAR The strong classifier of sign judges whether there is the secondary detection result of vehicle with the presence or absence of vehicle graphic in video image.

It is described to be based on video images detection in the vehicle checking method according to the present invention based on video and radar Judged when vehicle whether there is herein in connection with radar data, comprising the following steps: use the LBP operator extraction vehicle of 9 × 9 neighborhoods Textural characteristics obtain the strong classifier with LBP feature after being trained to sample image, using this with LBP feature Strong classifier judges whether there is the primary testing result of vehicle with the presence or absence of vehicle graphic in video image;In the primary When testing result detection has vehicle, using HAAR feature extraction vehicle textural characteristics, had after being trained to sample image The strong classifier of HAAR feature is judged in video image using the strong classifier with HAAR feature with the presence or absence of vehicle figure Whether shape is had the secondary detection result of vehicle and confidence level;Judged whether to be higher than default value according to confidence level, is to sentence Disconnected vehicle exists;Otherwise judge that vehicle whether there is based on radar data, be to judge that vehicle exists, otherwise judge that vehicle is not deposited ?.

In the vehicle checking method according to the present invention based on video and radar, the calculating vehicle detection index The step of include that following car speed calculates step: extract the automobile's instant velocity of vehicle based on radar data, judge that automobile's instant velocity is It is no to be higher than preset vehicle speed, it is that average speed is otherwise calculated based on video image and is made then using the automobile's instant velocity as car speed For car speed.

It is described to be calculated based on video image in the vehicle checking method according to the present invention based on video and radar Average speed is the average speed for being calculated by the following formula vehicle:

Wherein, D2 and D1 is respectively the corresponding reality of vehicles identifications horizontal line on nth frame and the 1st frame video image in video data The fore-and-aft distance of border road, T are the picture frame interval time of video image, VaUnit be kilometer per hour.

In the vehicle checking method according to the present invention based on video and radar, the calculating vehicle detection index The step of include that following queue length calculates step: the automobile's instant velocity of tail portion vehicle on section to be detected is judged based on radar data When lower than preset vehicle speed, video data is started based on using the method for telescopic window and calculates queue length, and is judged in radar data The moment speed of head vehicle on the section to be detected withdraws telescopic window when being greater than zero, cancels the queueing condition of vehicle.

The present invention also provides a kind of vehicle detecting system based on video and radar, comprising: radar installations, for acquiring The radar data of section vehicle to be detected;Video acquisition device, for acquiring the video image in section to be detected;Fusion detection dress It sets, is based on the radar data and video image, the vehicle graphic on video image is corresponding with vehicle on section to be detected Radar data matches, and calculates vehicle detection index;Display device, for the calculated result of vehicle detection index to be superimposed upon It is shown on video image.

Implement the vehicle checking method and system of the invention based on video and radar, has the advantages that this hair It is bright by blending video and detections of radar, can both realize radar data by superposition of data on the video images can Depending on changing, and it can use radar data and solve the problems, such as that video detection is affected by environment big, make detection data more comprehensively, detection Accuracy rate also greatly improves.

Detailed description of the invention

Fig. 1 is the flow chart according to the vehicle checking method based on video and radar of the preferred embodiment of the present invention;

Fig. 2 is the coordinate schematic diagram according to section to be detected on real road of the present invention;

Fig. 3 is the coordinate schematic diagram of the radar data acquired according to the present invention;

Fig. 4 is the video image in the section to be detected acquired according to the present invention;

Fig. 5 is the display situation map of Some vehicles Testing index in Fig. 4;

Fig. 6 is the vehicle holding time of the vehicle checking method according to the present invention based on video and radar to calculate step stream Cheng Tu;

Fig. 7 is the flow chart according to the present invention that whether there is step first embodiment based on video images detection vehicle;

Fig. 8 a and Fig. 8 b are respectively the example of positive sample image and negative sample image;

The present invention is based on the schematic diagrams of the LBP operator extraction vehicle textural characteristics of 9 × 9 neighborhoods according to Fig. 9;

Figure 10 is the flow chart according to the present invention that whether there is step second embodiment based on video images detection vehicle;

Figure 11 is that the car speed of the vehicle checking method according to the present invention based on video and radar calculates steps flow chart Figure;

The video image for the 1st frame and nth frame that Figure 12 a and 12b respectively choose according to the present invention;

The present invention is based on the module frame charts of the first embodiment of video and the vehicle detecting system of radar according to Figure 13;

The present invention is based on the module frames of the middle fusion detection device of the vehicle detecting system of video and radar according to Figure 14 Figure;

The present invention is based on the module frame charts of the second embodiment of video and the vehicle detecting system of radar according to Figure 15.

Specific embodiment

In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is A part of the embodiments of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.

Referring to Fig. 1, for according to the process of the vehicle checking method based on video and radar of the preferred embodiment of the present invention Figure.As shown in Figure 1, the embodiment provide the vehicle checking method based on video and radar the following steps are included:

The radar data of vehicle on S101, acquisition section to be detected.

The present invention installs radar installations on real road in advance, and delimits the square that the length along vehicle heading is L Shape region is section to be detected.As shown in Figure 2, start line L1 and terminated line L2 delimited on lane in the same direction, between the two structure The section to be detected for being L at length.Radar installations is mounted on the road edge in section to be detected, and with the installation of the radar installations Position acquires the speed and location information of vehicle in section to be detected as coordinate origin O.Referring to Fig. 3, for according to the present invention The coordinate schematic diagram of the radar data of acquisition.Coordinate origin O is radar installations infield in Fig. 3, which is that two dimension is sat Mark, i.e., radar installations and vehicle are indicated in the projection of ground level using polar mode.It is wide to respectively represent lane by X, Y in figure Spend direction and vehicle heading, VrIndicate the automobile's instant velocity of vehicle, rrIndicate vehicle location.VX,rxIndicate vehicle along lane The speed and location information of width direction, VY,ryIndicate vehicle along the speed and location information of vehicle heading, radar dress Set the speed that will test out the two directions and actual vehicle position.And be calculated by the following formula out vehicle automobile's instant velocity and Vehicle location:

S102, the video image for acquiring section to be detected.

The step can be realized by the video acquisition device being pre-installed on real road.According to referring to Fig. 4, The video image in the section to be detected that the present invention acquires.The present invention does not limit the execution sequence of above-mentioned steps S101 and S102 System can also first acquire video image, then acquire radar data, or both synchronous execute.

S103, it is based on radar data and video image, by the vehicle graphic on video image and vehicle on section to be detected Corresponding radar data matches, and calculates vehicle detection index.

Fig. 2 and Fig. 4 are please referred to, which detects vehicle graphic, such as C1 ' and C2 ' on the video images, and with The radar data of vehicle matches on section to be detected, such as by the automobile's instant velocity and practical vehicle of vehicle graphic C1 ' and vehicle C1 Position matches, and equally vehicle graphic C2 ' and vehicle C2 match.Subsequently, based on above-mentioned radar data and video image Calculate the vehicle detection index of each vehicle, including but not limited to environment visibility, vehicle holding time, vehicle location, vehicle The indexs such as road information, car speed, queue length, flow and/or road situation.

S104, it the calculated result of vehicle detection index is superimposed upon on video image shows.

As shown in figure 4, the car speed of vehicle and actual vehicle position are shown the vehicle figure on video image by the present invention Shape region.Specifically, the car speed of vehicle C1 and actual vehicle position are shown to the rectangle frame in vehicle graphic C1 ' In, as in Fig. 4 " P:20,13 " and " V:35 ".Preferably, the calculated result of Some vehicles Testing index can also be superimposed upon The upper left corner of video image is shown.The Some vehicles Testing index is particularly shown situation as shown in figure 5, aobvious in the upper left corner The vehicle detection index shown can include but is not limited to the flow of environment visibility and each lane, road situation, vehicle The indexs such as holding time and queue length.Wherein road situation can be expressed as congestion, slow and unimpeded three kinds of states. These vehicle detection indexs are the testing result in the measurement period of setting, can be updated with every 50ms primary.

The present invention can choose by the way that video image and radar data match and calculate effective vehicle detection and refer to Data are marked, and then are watched in real time on the video images.

In a preferred embodiment of the invention, vehicle data is matched by the affiliated lane of vehicle.Due to same a period of time It carves same lane and only exists a vehicle, therefore the present invention is based respectively on radar data and video images detection goes out the affiliated lane of vehicle Number judges the corresponding vehicle graphic detected of the radar data if number of track-lines is identical.

In the actual operation process, the present invention need in advance on the section to be detected of real road delimit start line L1 and Terminated line L2 establishes real road coordinate system as origin using the intersection point O of terminated line L2 and road edge.And measure each lane line Lateral distance XA at start line L11To XAn, i.e., each lane line LA on real road1To LAnIn real road coordinate system On abscissa numerical value.Then, the video image in the section to be detected is acquired, and start line L1 ' strokes and end on the video images Only area to be tested delimited between line L2 '.With the intersection point O ' of terminated line L2 ' on video image and road edge for origin, establish Video image coordinate system.Lateral distance XA of each lane line at start line L1 ' in area to be tested is measured simultaneously1' extremely XAn', i.e., each lane line LA on video image1' to LAn' abscissa on video image coordinate system.

In the vehicle checking method based on video and radar, above-mentioned steps S103 by video image vehicle with it is corresponding Radar data the step of matching specifically include:

It is to obtain vehicle graphic to exist firstly, whether there is vehicle graphic in identification area to be tested on the video images Vehicle graphic coordinate (r on video imagex', ry').The vehicle graphic coordinate includes abscissa rx' and ordinate ry'.With On the basis of start line L1 ' on video image, start to detect the affiliated lane of vehicle when vehicle enters the start line position L1 '.Root Process according to the affiliated lane information of video images detection vehicle graphic is as follows:

When according to ordinate ry' judge that the vehicle graphic is in the start line L1 ' delimited in advance in the area to be tested When upper, i.e. ryWhen '=L, according to the abscissa r of vehicle graphicx' and the area to be tested of measured in advance on each lane line Lateral distance XA1' to XAn' judge the vehicle graphic lane information, if XAi’≤|rx’|<XAi+1', then judge the vehicle figure Shape belongs to i-th of lane, 1≤i < n;Wherein n is the quantity in lane in area to be tested.As determined vehicle graphic C2 ' in Fig. 4 Abscissa meets XA2’≤rx’≤XA3', therefore, it is determined that vehicle graphic C2 ' is located at the 2nd lane.

Actual position coordinate (r of the vehicle in real road coordinate system is extracted subsequently, based on radar datax, ry), the reality Border position coordinates include abscissa rxAnd ordinate ry.On the basis of the start line L1 on real road, enter starting in vehicle Start to detect vehicle affiliated lane when the line position L1.The process for detecting the affiliated lane information of vehicle according to radar data is as follows:

When according to ordinate ryWhen judging that the vehicle is on the start line L1 delimited in advance in section to be detected, i.e. ry=L When, according to the abscissa r of vehiclexAnd on the section to be detected of measured in advance each lane line lateral distance LA1To LAnSentence Break the lane information of the vehicle, if LAj≤|rx|<LAj+1, then judge that the vehicle belongs to j-th of lane, 1≤j < n;Wherein n is The quantity in lane on section to be detected.As determined in Fig. 2, the abscissa of vehicle graphic C2 meets XA2≤rx≤XA3, therefore, it is determined that Vehicle C2 is located at the 2nd lane.

Finally, judge whether number of track-lines based on video images detection and the number of track-lines detected based on radar data are identical, As i=j, the radar data of the vehicle graphic and the vehicle that will identify that matches;Otherwise it mismatches.For example, aforementioned vehicle Figure C2 ' and vehicle C2 is respectively positioned on the 2nd lane, therefore can be by the automobile's instant velocity and practical vehicle of vehicle graphic C2 ' and vehicle C2 Position matches.

The present invention also detects the vehicle holding time.Referring to Fig. 6, being based on video and radar to be according to the present invention Vehicle checking method the vehicle holding time calculate flow chart of steps.As shown in fig. 6, in a preferred embodiment of the present invention In, aforementioned calculating vehicle detection index specifically includes following vehicle holding time calculating step:

S201, process start;

Whether S202, detection environment visibility are greater than default visibility, are to go to step S208-S212 to start based on video Image detection vehicle whether there is, and record the vehicle holding time, otherwise goes to step S203-S207 and is based on radar data detection vehicle It whether there is, and record the vehicle holding time.The default visibility is preferably 40%-60%, such as judges environment visibility Whether 50% is greater than.

S203, it has detected whether that vehicle enters section to be detected based on radar data, has been to go to step S204, otherwise turns to walk Rapid S213 terminates process;It can judge whether vehicle is located at road to be detected by the actual vehicle position of detections of radar in the step In section.

S204, record vehicle entry time T1;

S205, it has detected whether that vehicle leaves detection zone based on radar data, has been to go to step S206, otherwise goes to step S205 is continued to test;

S206, record vehicle time departure T2;

S207, calculating vehicle holding time are T2-T1;

S208, whether there is vehicle to enter area to be tested based on video images detection, be to go to step S209, otherwise turn to walk Rapid S213 terminates process;

S209, record vehicle entry time T3;

S210, whether there is vehicle to leave area to be tested based on video images detection, be to go to step S211, otherwise turn to walk Rapid S210 is continued to test;

S211, record vehicle time departure T4;

S212, calculating vehicle holding time are T4-T3;

S213, process terminate.

The detailed process that whether there is in the present invention based on video images detection vehicle is described below.The present invention adopts Vehicle graphic is detected and is identified with the second level vehicle feature recognition algorithm of Machine self-learning.Due to road, vehicle with And the complexity of weather environment, current many video frequency vehicle recognizers, including background subtraction, Multi Frame Difference point-score, light stream Method, part detection, feature detection etc. can be all subject to certain restrictions.Present invention employs the local vehicle characteristic methods of two-stage, i.e., Extensive primary detection is carried out using follow-on LBP feature, and two then are carried out using HAAR feature to primary testing result Secondary identification, output have the result of recognition confidence.

Referring to Fig. 7, whether there is step first embodiment based on video images detection vehicle to be according to the present invention Flow chart.As shown in fig. 7, in above-mentioned steps S208-S212 based on video images detection vehicle with the presence or absence of the step of include:

S301, process start;

S302, the LBP operator extraction vehicle textural characteristics using 9 × 9 neighborhoods, obtain band after being trained to sample image There is the strong classifier of LBP feature, is judged in video image using the strong classifier with LBP feature with the presence or absence of vehicle figure Whether shape is had the primary testing result of vehicle.S303 is gone to step when the detection of primary testing result has vehicle further to analyze, it is no It then goes to step S305 and judges that vehicle is not present.

The resolution ratio for the video image that the present invention acquires is 1080P, and the minimum pixel requirement of the vehicle graphic detected is 100 or so, therefore the present invention is extended the LBP operator of traditional 3 × 3 neighborhoods, uses the LBP operator extraction of 9 × 9 neighborhoods Vehicle textural characteristics.Therefore, the specific mistake of extensive primary detection is carried out in step S302 using follow-on LBP feature Journey is as follows:

Firstly, the exemplary positive sample image of Fig. 8 a and the exemplary negative sample image of Fig. 8 b are normalized the present invention, rule Size of the model to 64*64.

Image texture is extracted subsequently, based on positive and negative sample image.The image of LBP operator extraction based on traditional 3 × 3 neighborhoods Texture T can be used basic operation unit and be defined: entire window is by a center window gcWith 8 adjacent 3 × 3 pixels Child window g0,......,g7Composition, is shown below:

T~(g0-gc,g1-gc,….g7-gc);

9 × 9 pixel basic operation units after extension are then divided into 93 × 3 pixel child windows by the present invention, such as Fig. 9 institute Show.The average gray of 9 child windows is sought respectively, and original 3 × 3 child window is replaced with this average value, extracts image line Manage T '.Image texture T ' may be expressed as:

T~(g0’-gc’,g1’-gc’,….g7’-gc');

By this simple operation, on the one hand 1/9 that compression of images is original image can be greatly reduced the meter of subsequent algorithm Calculation amount, and also comply with actual vehicle pixel request.On the other hand 9 × 9 neighborhood extending LBP operators can be transformed to basic 3 × 3LBP operation operator.

Then with center pixel gcFor reference pixel, local binarization processing is carried out to rest of pixels in window.Processing side Method may be expressed as:

Image texture characteristic after binarization operation may be expressed as:

Ts~(S(g0’-gc’),S(g1’-gc’),….S(g7’-gc’))

At this point, 8 binary numbers can be used to indicate in basic operation unit, pixel is weighted using following formula and is asked And processing, basic operation unit LBP characteristic value can be obtained are as follows:

The LBP characteristic value describes edge, texture, characteristic point and the plane domain of basic operation unit regional area.It will Aforesaid operations method expands to entire sample image region, and binary sequence description and the LBP characteristic pattern of sample image can be obtained.

Based on above modified LBP characterization method, passed through using discrete adaboost machine learning algorithm to a large amount of The training of the negative sample image of the positive sample image and some specific environments of vehicle selects most general character and identifies energy The LBP feature of power, and based on this characteristic component classifier, construct corresponding Weak Classifier.Then, grade is carried out to weak typing Connection, composition one has the strong classifier of LBP feature.Utilize the to be detected of the strong classifier judgement acquisition with LBP feature It whether there is vehicle graphic in the video image in section, and whether had the primary testing result of vehicle.

S303, using HAAR feature extraction vehicle textural characteristics, obtained after being trained to sample image special with HAAR The strong classifier of sign judges to obtain in video image with the presence or absence of vehicle graphic using the strong classifier with HAAR feature Whether the secondary detection result of vehicle is had.S304 is gone to step when primary testing result is detected with the presence of vehicle and judges vehicle, is otherwise turned Step S305 judges that vehicle is not present.

Using HAAR feature to be recognized primary testing result in step S303, detailed process is as follows:

Using the vehicle's contour and textural characteristics of the positive and negative sample image of HAAR feature extraction, by real-time AdaBoost machine Learning algorithm is trained obtained HAAR feature, and building second level has the strong classifier of confidence level.4 kinds should be used in the process Type, totally 6 kinds of HAAR feature composition characteristic templates, carry out edge to vehicle target and textural characteristics describe, including 2 kinds of edge spies Sign, 2 kinds of linear characters, a kind to corner characteristics and a kind of central feature.HAAR characterizing definition is that chequered with black and white rectangle is divided, can be with Carry out quantitative expression using the gray scale difference in image in adjacent rectangle region, both using white rectangle pixel in image and subtracting black Rectangular pixels and.In order to preferably extract characteristics of image, the present invention has carried out gray processing and histogram equalization to positive sample image Change processing.

It whether there is vehicle graphic in video image using the obtained strong classifier judgement acquisition with HAAR feature, Whether the secondary detection of vehicle is had as a result, and using the secondary detection result as final testing result.

S304, judge that vehicle exists.

S305, judge that vehicle is not present.

S306, process terminate.

Referring to Fig. 10, whether there is step second embodiment based on video images detection vehicle to be according to the present invention Flow chart.As shown in Figure 10, the second embodiment is essentially identical with first embodiment, the difference is that only, is being based on video Image detection vehicle judged while whether there is herein in connection with radar data, specifically includes the following steps:

S401, process start;

S402, the LBP operator extraction vehicle textural characteristics using 9 × 9 neighborhoods, obtain band after being trained to sample image There is the strong classifier of LBP feature, is judged in video image using the strong classifier with LBP feature with the presence or absence of vehicle figure Whether shape is had the primary testing result of vehicle.S403 is gone to step when the detection of primary testing result has vehicle further to analyze, it is no It then goes to step S407 and judges that vehicle is not present.

S403, using HAAR feature extraction vehicle textural characteristics, obtained after being trained to sample image special with HAAR The strong classifier of sign judges to obtain in video image with the presence or absence of vehicle graphic using the strong classifier with HAAR feature Whether the secondary detection result of vehicle and confidence level is had.S404 is gone to step when the detection of secondary detection result has vehicle to carry out in next step Judgement, otherwise goes to step S407 and judges that vehicle is not present.

The HAAR feature that the present invention uses has many advantages, such as that recognition accuracy is high, robustness is good, special compared to training LBP The binaryzation discrete type AdaBoost of sign, can be trained within the scope of entire gray values, so its obtain strong point Appliances have the predicted value size of testing result, i.e. confidence level.Therefore it is defeated that the detection of HAAR feature is utilized in the second embodiment Confidence level parameter out is analyzed and is compared to the authenticity for obtaining result convenient for subsequent.

S404, judged whether to be higher than default value according to confidence level, be to go to step S406 to judge that vehicle exists, otherwise turn Step S405 carries out next step judgement.For example, setting default value as 50%.When step S403 output confidence level be greater than 50%, It then can be confirmed to be real vehicle target;If it is less than 50%, it is likely that be error detection, it is possible to not be real vehicle Target, this when are corresponding with the target of the detections of radar in this lane again.

S405, judge that vehicle whether there is based on radar data, be to go to step S406 to judge that vehicle exists, otherwise turn to walk Rapid S407 judges that vehicle is not present.That is, can be confirmed to be yes if radar data also detects that the target of the vehicle Real target, otherwise it is assumed that being false vehicle target.Therefore, the present invention is detecting after carrying out secondary detection to video image To under vehicle graphic and the lower situation of confidence level, it can further be judged by radar data, improve the standard of vehicle detection True property.

S406, judge that vehicle exists.

S407, judge that vehicle is not present.

S408, process terminate.

Also car speed is detected in the present invention.Figure 11 is please referred to, is based on video and thunder to be according to the present invention The car speed of the vehicle checking method reached calculates flow chart of steps.As shown in figure 11, in a preferred embodiment of the present invention In, aforementioned calculating vehicle detection index specifically includes following car speed and calculates step:

S501, process start;

S502, the automobile's instant velocity V that vehicle is extracted based on radar datar, judge automobile's instant velocity VrWhether preset vehicle speed is higher than, It is to go to step S503, otherwise goes to step S504;

S503, the automobile's instant velocity V for arriving detections of radarrAs car speed V;

S504, average speed V is calculated based on video imageaAs car speed V.

S505, process terminate.

By taking preset vehicle speed is 5km/h as an example, work as VrThen car speed V is the average speed V of video detection when≤5a;Otherwise Car speed V value is the automobile's instant velocity V of detections of radarr.Since radar velocity measurement is using Doppler frequency shift principle, measurement It is the relative velocity between object and radar, measuring speed accuracy is significantly if car speed is relatively low or stationary It reduces, so the present invention will combine radar velocity measurement and video frequency speed-measuring, to improve tachometric survey accuracy.

Preferably, the present invention is based in the vehicle checking method of video and radar, average vehicle is calculated based on video image Fast VaDetailed process is as follows.The average speed V of vehicle of the present inventionaIt can be according to the location information displacement on the image of vehicle To calculate.Firstly, detecting vehicle graphic on the 1st frame and the video image of nth frame.And vehicle mark is drawn on the vehicle graphic Horizontal line is known, as depicted the 1st frame vehicles identifications horizontal line S2 and nth frame vehicles identifications horizontal line S1 in Figure 12 a and Figure 12 b respectively. Calculate the fore-and-aft distance D2 and D1 on the corresponding real road of vehicles identifications horizontal line along vehicle heading.

It is calculated by the following formula the average speed of vehicle:

Wherein, T is the picture frame interval time of video image, VaUnit be kilometer per hour.Preferably, image interframe 40ms is chosen every time T.

The velocity amplitude that the present invention is measured by the above method, is the actual average speed value of vehicle, and requires no complexity Image calibration and coordinate transform.

The present invention also detects the queue length of vehicle.It is the method using telescopic window in the present invention to video image It is detected, to calculate vehicle queue length.The method of telescopic window is to judge vehicle based on the method for morphologic edge extracting Presence, cardinal principle is: the sum of several row grey scale pixel values are greater than a certain threshold value, and telescopic window at the top of virtual region Interior whole grey scale pixel value and and previous moment window in whole pixels when detesting angle value and its difference less than a certain threshold value, then telescopic window is stretched A long pixel unit.Conversely, telescopic window shortens a pixel unit.

If the information of vehicles that edge detection obtains is incomplete, this just brings very big be stranded to the setting of telescopic window threshold value Difficulty, has vehicle queue in practical application but the phenomenon that telescopic window can not extend, and causes few detection.If when vehicle is no longer lined up When starting to run at a low speed, i.e., only wagon flow without queue in the case where, telescopic window cannot be retracted rapidly, in such a scenario Still dequeue is detected, error detection is caused.Therefore the characteristics of present invention can accurately detect moving target using radar data, it is first First parking is lined up and does a pre- judgement.

In a preferred embodiment of the invention, it is long to specifically include following vehicle queue for aforementioned calculating vehicle detection index Degree calculates step: judging whether the automobile's instant velocity of tail portion vehicle on section to be detected is lower than preset vehicle speed based on radar data, is It then starts based on video data and calculates queue length using the method for telescopic window, otherwise do not operate.The preset vehicle speed is preferably 5km/h.I.e. when vehicle is by the smooth queuing to, later vehicle can be by fast in the tail portion of area to be tested, car speed Slack-off, the automobile's instant velocity that radar installations detects also can slowly be reduced to 5 kilometers/hour hereinafter, starting to carry out video row at this time The detection of team's algorithm, using the marginal information of vehicle, gradient difference and color difference etc. obtain the accurate information of vehicle queue.

Judge whether the moment speed of the head vehicle on section to be detected is greater than zero subsequently, based on radar data, is then Telescopic window is withdrawn, cancels the queueing condition of vehicle, does not otherwise operate.I.e. when vehicle by be lined up begin to change into be not lined up when, to When the head vehicle of detection zone starts mobile, radar installations meeting accurate detection is arrived, and at this time withdraws telescopic window, is cancelled The queueing condition of vehicle.

Please refer to Figure 13, according to the present invention is based on the modules of the first embodiment of video and the vehicle detecting system of radar Block diagram.As shown in figure 13, the vehicle detecting system based on video and radar which provides includes: radar installations 10, view Frequency acquisition device 20, fusion detection device 30 and display device 40.

Wherein radar installations 10 is used to acquire the radar data of section vehicle to be detected.Radar installations 10 of the invention is installed Road edge in section to be detected.

Video acquisition device 20 is used to acquire the video image in section to be detected.The video acquisition device 20 can be using peace Camera on section to be detected is realized.

Fusion detection device 30 is connect with radar installations 10 and video acquisition device 20, for being based on radar data and video Vehicle graphic radar data corresponding with vehicle on section to be detected on video image is matched, and calculates vehicle by image Testing index.

Display device 40 is connected with detection device 30 is merged, for the calculated result of vehicle detection index to be superimposed upon video It is shown on image, convenient for viewing detection effect in real time.Preferably, which can use through Ethernet and melt Close the semaphore and signal lamp realization that detection device 30 connects.

It includes but is not limited to that environment visibility, vehicle account for that the vehicle detection index that detection device 30 calculates is merged in the present invention Having time, vehicle location, lane information, car speed, queue length, flow and/or road situation.

Please refer to Figure 14, according to the present invention is based on the middle fusion detection devices 30 of the vehicle detecting system of video and radar Module frame chart.As shown in figure 14, fusion detection device 30 includes but is not limited to one or more following unit: vehicle exists Matching unit 31, vehicle holding time computing unit 32, car speed computing unit 33 and queue length computing unit 34.

Wherein there are matching units 31 to be used for the vehicle graphic on video image and vehicle pair on section to be detected for vehicle The radar data answered matches.Specifically, the vehicle there are matching unit 31 be based on video image identification area to be tested in be No there are vehicle graphics, are to obtain the vehicle graphic coordinate of vehicle graphic on the video images, the vehicle graphic coordinate packet Include abscissa rx' and ordinate ry', when according to ordinate ry' judge that the vehicle graphic is in the area to be tested preparatory When in the start line of delimitation, according to the abscissa r of vehicle graphicx' and the area to be tested of measured in advance on each lane line Lateral distance XA1' to XAn' judge the vehicle graphic lane information, if XAi’≤|rx’|<XAi+1', then judge the vehicle Figure belongs to i-th of lane, 1≤i < n;Wherein n is the quantity in lane in area to be tested.It is extracted subsequently, based on radar data The actual position coordinate of vehicle, the actual position coordinate include abscissa rxAnd ordinate ry, when according to ordinate ryJudgement When the vehicle is in the start line delimited in advance in section to be detected, according to the abscissa r of vehiclexAnd measured in advance to Detect the lateral distance XA of each lane line on section1To XAnThe lane information of the vehicle is judged, if XAj≤|rx|<XAj+1, Then judge that the vehicle belongs to j-th of lane, 1≤j < n;Wherein n is the quantity in lane on section to be detected.As i=j, will know Not Chu the radar data of vehicle graphic and the vehicle match.

Vehicle holding time computing unit 32 is for calculating the vehicle holding time.Firstly, calculating environment based on video image Visibility.Judge whether environment visibility is higher than default visibility again, be, whether there is based on video images detection vehicle, and Record the holding time;Otherwise the vehicle holding time is detected based on radar data.

Preferably, vehicle holding time computing unit 32 is based on whether video images detection vehicle is deposited by following steps : 1) the LBP operator extraction vehicle textural characteristics of 9 × 9 neighborhoods are used, are obtained after being trained to sample image special with LBP The strong classifier of sign judges to whether there is vehicle graphic in video image, is using the strong classifier with LBP feature The no primary testing result for having vehicle.2) when the primary testing result detection has vehicle, HAAR feature extraction vehicle texture is used Feature obtains the strong classifier with HAAR feature after being trained to sample image, utilize strong point with HAAR feature Class device judges whether there is the secondary detection result of vehicle with the presence or absence of vehicle graphic in video image.

It is highly preferred that vehicle holding time computing unit 32 is based on whether video images detection vehicle is deposited by following steps : 1) the LBP operator extraction vehicle textural characteristics of 9 × 9 neighborhoods are used, are obtained after being trained to sample image special with LBP The strong classifier of sign judges to whether there is vehicle graphic in video image, is using the strong classifier with LBP feature The no primary testing result for having vehicle.2) when the detection of primary testing result has vehicle, using HAAR feature extraction vehicle textural characteristics, The strong classifier with HAAR feature is obtained after being trained to sample image, is sentenced using the strong classifier with HAAR feature It whether there is vehicle graphic in disconnected video image, whether there is the secondary detection result of vehicle and confidence level.3) according to confidence Degree judges whether to be higher than default value, is to judge that vehicle exists;Otherwise judge that vehicle whether there is based on radar data, be then Judge that vehicle exists, otherwise judges that vehicle is not present.

Car speed computing unit 33 is for calculating car speed.Firstly, extracting the instantaneous vehicle of vehicle based on radar data Speed, judges whether automobile's instant velocity is higher than preset vehicle speed, is otherwise to be based on video figure then using the automobile's instant velocity as car speed As calculating average speed as car speed.

Preferably, which is calculated by the following formula the average speed of vehicle:

Wherein, D2 and D1 is respectively the corresponding reality of vehicles identifications horizontal line on nth frame and the 1st frame video image in video data The fore-and-aft distance of border road, T are the picture frame interval time of video image, VaUnit be kilometer per hour.

Queue length computing unit 34 is for calculating queue length.Firstly, being judged on section to be detected based on radar data When the automobile's instant velocity of tail portion vehicle is lower than preset vehicle speed, starts based on video data and calculate to be lined up using the method for telescopic window and grow Degree, and telescopic window is withdrawn when radar data judges the moment speed of the head vehicle on the section to be detected greater than zero, it takes Disappear the queueing condition of vehicle.

Please refer to Figure 15, according to the present invention is based on the modules of the second embodiment of video and the vehicle detecting system of radar Block diagram.As shown in figure 15, the vehicle detecting system based on video and radar and first embodiment that second embodiment provides are basic It is identical, the difference is that, it is additionally arranged radar processing board 50 and video processing board 60.

Wherein, radar processing board 50 and radar installations 10 communicate, and obtain radar data and are handled.Radar processing board 50 It can be whether there is based on radar data detection vehicle, vehicle holding time, vehicle location, the real-time indicators such as automobile's instant velocity.

Video processing board 60 and video acquisition device 20 communicate, and obtain video data and are handled.Video processing board 60 It can be whether there is based on video data detection environment visibility, vehicle, vehicle holding time, the real-time indicators such as queue length.

Fusion detection device 30 is connected with the radar processing board 50 and video processing board 60, main to pass through fusion radar processing The real-time detector data of plate 50 and video processing board 60, the vehicle graphic on video image is corresponding with vehicle on section to be detected Radar data match, and obtain the calculated value of effective vehicle detection index.Wherein radar processing board 50 can by with Too net is communicated with detection device 30 is merged.Video processing board 60 can use but be not limited to RS232 interface with merge detection device 30 communications.Vehicle whether there is, vehicle holding time, lane information, car speed, congestion in road situation and flow wait vehicles Testing index, which can be according to the calculation method of aforementioned each vehicle detection index, at radar The calculated result that reason plate 50 and video processing board 60 obtain, which is chosen, perhaps calculates effective data or the statistics week in setting Related data is counted in phase.For only have radar data or video data could calculated vehicle detection index, such as Based on the calculated vehicle location of radar data, and it is based on the calculated environment visibility of video data and queue length, then Directly the calculated result of radar processing board 50 or video processing board 60 can be exported.

It should be appreciated that the principle of the vehicle detecting system and method for the invention based on video and radar and realizing Cheng Xiangtong, thus to the embodiment based on video and the vehicle checking method of radar elaborate be also applied for based on video and The vehicle detecting system of radar.

In conclusion the present invention uses two kinds of non-contact type vehicle detecting sensors of video and radar, on hardware, software It is completely integrated, constitute a kind of really round-the-clock, visualization, comprehensive Testing index, high detection rate, high reliability New vehicle detection system.Most basic detection data is provided to solve Urban Traffic Jam Based.The vehicle detecting system exists The respective advantage that video and radar sensor are combined on hardware device, solving radar monitoring data cannot visually ask The problem of topic, it is affected by environment also to compensate for video detection, is unable to all weather operations, has also merged two kinds of biographies in software algorithm The advantage of sensor, so that the accuracy rate of vehicle detection also greatly improves.

Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features; And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and Range.

Claims (9)

1. a kind of vehicle checking method based on video and radar, which comprises the following steps:
Acquire the radar data of section vehicle to be detected;
Acquire the video image in section to be detected;
It is based on the radar data and video image, the vehicle graphic on video image is corresponding with vehicle on section to be detected Radar data matches, and calculates vehicle detection index;
The calculated result of vehicle detection index is superimposed upon on video image and is shown;
Wherein, the step that the vehicle graphic by video image radar data corresponding with vehicle on section to be detected matches Suddenly include:
It is to obtain vehicle graphic on the video images based on whether there is vehicle graphic in video image identification area to be tested Vehicle graphic coordinate, the vehicle graphic coordinate includes abscissa rx' and ordinate ry', when according to ordinate ry' judgement When the vehicle graphic is in the start line delimited in advance in the area to be tested, according to the abscissa r of vehicle graphicx' with And in the area to be tested of measured in advance each lane line lateral distance XA1' to XAn' judge the vehicle graphic lane information, If XAi’≤|rx’|<XAi+1', then judge that the vehicle graphic belongs to i-th of lane, 1≤i < n;Wherein n is in area to be tested The quantity in lane;
The actual position coordinate of vehicle is extracted based on radar data, the actual position coordinate includes abscissa rxAnd ordinate ry, when according to ordinate ryWhen judging that the vehicle is in the start line delimited in advance in section to be detected, according to the horizontal seat of vehicle Mark rxAnd on the section to be detected of measured in advance each lane line lateral distance XA1To XAnJudge the lane letter of the vehicle Breath, if XAj≤|rx|<XAj+1, then judge that the vehicle belongs to j-th of lane, 1≤j < n;Wherein n is lane on section to be detected Quantity;
As i=j, the radar data of the vehicle graphic and the vehicle that will identify that matches.
2. the vehicle checking method according to claim 1 based on video and radar, which is characterized in that the vehicle detection Index include environment visibility, the vehicle holding time, vehicle location, lane information, car speed, queue length, flow and/or Road situation.
3. the vehicle checking method according to claim 2 based on video and radar, which is characterized in that the calculating vehicle Include the following vehicle holding time calculating step in the step of Testing index:
Environment visibility is calculated based on video image;
Judge whether environment visibility is higher than default visibility, be, whether there is based on video images detection vehicle, and record Holding time;Otherwise the vehicle holding time is detected based on radar data.
4. the vehicle checking method according to claim 3 based on video and radar, which is characterized in that described to be based on video Image detection vehicle with the presence or absence of the step of include:
Using the LBP operator extraction vehicle textural characteristics of 9 × 9 neighborhoods, obtained after being trained to sample image with LBP feature Strong classifier, using this with LBP feature strong classifier judge in video image whether there is vehicle graphic, obtain whether There is the primary testing result of vehicle;
When the primary testing result detection has vehicle, using HAAR feature extraction vehicle textural characteristics, sample image is carried out The strong classifier with HAAR feature is obtained after training, judged using the strong classifier with HAAR feature be in video image No there are vehicle graphics, whether are there is the secondary detection result of vehicle.
5. the vehicle checking method according to claim 3 based on video and radar, which is characterized in that described to be based on video Image detection vehicle is judged when whether there is herein in connection with radar data, comprising the following steps:
Using the LBP operator extraction vehicle textural characteristics of 9 × 9 neighborhoods, obtained after being trained to sample image with LBP feature Strong classifier, using this with LBP feature strong classifier judge in video image whether there is vehicle graphic, obtain whether There is the primary testing result of vehicle;
When the primary testing result detection has vehicle, using HAAR feature extraction vehicle textural characteristics, sample image is carried out The strong classifier with HAAR feature is obtained after training, judged using the strong classifier with HAAR feature be in video image No there are vehicle graphics, whether are there is the secondary detection result of vehicle and confidence level;
Judge whether confidence level is higher than default value, be, judges that vehicle exists;Otherwise whether vehicle is judged based on radar data In the presence of, be judge vehicle exist, otherwise judge that vehicle is not present.
6. the vehicle checking method according to claim 2 based on video and radar, which is characterized in that the calculating vehicle The step of Testing index includes that following car speed calculates step:
The automobile's instant velocity that vehicle is extracted based on radar data, is judged whether automobile's instant velocity is higher than preset vehicle speed, is then by the wink When speed as car speed, otherwise based on video image calculate average speed as car speed.
7. the vehicle checking method according to claim 6 based on video and radar, which is characterized in that described to be based on video It is the average speed for being calculated by the following formula vehicle that image, which calculates average speed:
Wherein, D2 and D1 is respectively the corresponding practical road of vehicles identifications horizontal line on nth frame and the 1st frame video image in video data The fore-and-aft distance on road, T are the picture frame interval time of video image, VaUnit be kilometer per hour.
8. the vehicle checking method according to claim 2 based on video and radar, which is characterized in that the calculating vehicle The step of Testing index includes that following queue length calculates step:
When judging the automobile's instant velocity of tail portion vehicle on section to be detected lower than preset vehicle speed based on radar data, video is started based on Data calculate queue length using the method for telescopic window, and judge the head vehicle on the section to be detected in radar data Moment speed withdraws telescopic window when being greater than zero, cancels the queueing condition of vehicle.
9. a kind of vehicle detecting system based on video and radar characterized by comprising
Radar installations, for acquiring the radar data of section vehicle to be detected;
Video acquisition device, for acquiring the video image in section to be detected;
Merge detection device, be based on the radar data and video image, by video image vehicle graphic and road to be detected The corresponding radar data of vehicle matches in section, and calculates vehicle detection index;
Display device is shown for the calculated result of vehicle detection index to be superimposed upon on video image;
The fusion detection device includes that there are matching units for vehicle, for performing the following operations:
It is to obtain vehicle graphic on the video images based on whether there is vehicle graphic in video image identification area to be tested Vehicle graphic coordinate, the vehicle graphic coordinate includes abscissa rx' and ordinate ry', when according to ordinate ry' judgement When the vehicle graphic is in the start line delimited in advance in the area to be tested, according to the abscissa r of vehicle graphicx' with And in the area to be tested of measured in advance each lane line lateral distance XA1' to XAn' judge the vehicle graphic lane information, If XAi’≤|rx’|<XAi+1', then judge that the vehicle graphic belongs to i-th of lane, 1≤i < n;Wherein n is in area to be tested The quantity in lane;
The actual position coordinate of vehicle is extracted based on radar data, the actual position coordinate includes abscissa rxAnd ordinate ry, when according to ordinate ryWhen judging that the vehicle is in the start line delimited in advance in section to be detected, according to the horizontal seat of vehicle Mark rxAnd on the section to be detected of measured in advance each lane line lateral distance XA1To XAnJudge the lane letter of the vehicle Breath, if XAj≤|rx|<XAj+1, then judge that the vehicle belongs to j-th of lane, 1≤j < n;Wherein n is lane on section to be detected Quantity;
As i=j, the radar data of the vehicle graphic and the vehicle that will identify that matches.
CN201610816421.0A 2016-09-12 2016-09-12 A kind of vehicle checking method and system based on video and radar CN106373394B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610816421.0A CN106373394B (en) 2016-09-12 2016-09-12 A kind of vehicle checking method and system based on video and radar

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610816421.0A CN106373394B (en) 2016-09-12 2016-09-12 A kind of vehicle checking method and system based on video and radar

Publications (2)

Publication Number Publication Date
CN106373394A CN106373394A (en) 2017-02-01
CN106373394B true CN106373394B (en) 2019-01-04

Family

ID=57899336

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610816421.0A CN106373394B (en) 2016-09-12 2016-09-12 A kind of vehicle checking method and system based on video and radar

Country Status (1)

Country Link
CN (1) CN106373394B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106710240A (en) * 2017-03-02 2017-05-24 公安部交通管理科学研究所 Passing vehicle tracking and speed measuring method integrating multiple-target radar and video information
CN108337785A (en) * 2018-02-07 2018-07-27 宁夏乐智博科技有限公司 Illumination control method and intelligent illumination system
CN108200708A (en) * 2018-02-07 2018-06-22 宁夏乐智博科技有限公司 Control method and smart streetlamp
CN108337786A (en) * 2018-02-07 2018-07-27 宁夏乐智博科技有限公司 Illumination control method and smart streetlamp

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3655496B2 (en) * 1999-06-07 2005-06-02 三菱電機株式会社 Vehicle detection device and a vehicle detection method and a computer readable recording medium recording the vehicle detection program
KR101030763B1 (en) * 2010-10-01 2011-04-26 위재영 Image acquisition unit, acquisition method and associated control unit
JP5597746B2 (en) * 2013-05-15 2014-10-01 三菱電機株式会社 Vehicle detection system, the vehicle detection method of the vehicle detection device and a vehicle detection system
CN104424804B (en) * 2013-09-10 2016-08-10 上海弘视通信技术有限公司 Intelligent multi-lane single radar gun method and system
CN103581627A (en) * 2013-11-07 2014-02-12 北京环境特性研究所 Image and information fusion display method for high-definition video
CN105480227B (en) * 2015-12-29 2017-12-15 大连楼兰科技股份有限公司 Active driving technology based on radar and infrared video image information fusion

Also Published As

Publication number Publication date
CN106373394A (en) 2017-02-01

Similar Documents

Publication Publication Date Title
Zhang et al. Data-driven intelligent transportation systems: A survey
US7804980B2 (en) Environment recognition device
Coifman et al. A real-time computer vision system for vehicle tracking and traffic surveillance
EP0454166B1 (en) Traffic flow measuring method and apparatus
US20100021009A1 (en) Method for moving targets tracking and number counting
Hautiére et al. Automatic fog detection and estimation of visibility distance through use of an onboard camera
US9472097B2 (en) Roadway sensing systems
Huang et al. On-board vision system for lane recognition and front-vehicle detection to enhance driver's awareness
Beymer et al. A real-time computer vision system for measuring traffic parameters
EP0986036A2 (en) Method of updating reference background image, method of detecting entering objects and system for detecting entering objects using the methods
US5554983A (en) Object recognition system and abnormality detection system using image processing
CN1980322B (en) Object detecting system and object detecting method
JP3669205B2 (en) Obstacle recognition device
US20020054210A1 (en) Method and apparatus for traffic light violation prediction and control
US8849554B2 (en) Hybrid traffic system and associated method
CN100538763C (en) Video based mixed traffic flow parameter detecting method
US9824277B2 (en) Pedestrian right of way monitoring and reporting system and method
WO2000077736A1 (en) Method of detecting objects within a wide range of a road vehicle
CN101750049B (en) Distance measurement method is based on the monocular vision road car and the vehicle itself information
CN102542797B (en) Image-based detecting system and method for traffic parameters and computer program product thereof
CN102165493B (en) Detection of vehicles in an image
CN1984236A (en) Method for collecting characteristics in telecommunication flow information video detection
CN101710448B (en) Road traffic state detecting device based on omnibearing computer vision
CN1223964C (en) Apparatus and method for measuring vehicle queue length
US6188778B1 (en) Traffic congestion measuring method and apparatus and image processing method and apparatus

Legal Events

Date Code Title Description
C06 Publication
SE01
GR01