CN103809163A - Local maximum value based vehicle radar target detection method - Google Patents

Abstract

The invention discloses a local maximum value based vehicle radar target detection method. The local maximum value based vehicle radar target detection method includes the following steps that step one, a microwave detection device on the road surface uploads collected echoes to an upper computer through a serial port, radar echo distribution conditions of the road surface are observed through the upper computer, and radar echo areas are partitioned; step two, the range of a search window is calculated according to a distance resolution ratio of frequency modulation continuous wave (FMCW) radar and width information of a lane; step three, all lane echo points extracted in the step one are traversed, the search window in the step two is set in the neighborhood of each echo point for each echo point, and subsequently, the echo point is compared with other echo points in the search window; step four, whether an echo intensity value of a searched local maximum value point is larger than a set echo threshold is judged, the searched local maximum value point is a target point if the searched local maximum value point is larger than the set echo threshold, and otherwise, the searched local maximum value point is considered to be a non-target point; step five, the target point is marked, and an echo signal is turned into a two-value signal so as to achieve target detection.

Description

A kind of Radar for vehicle object detection method based on local maximum

Technical field

The present invention relates to a kind of method of Radar Targets'Detection, particularly a kind of Radar for vehicle object detection method based on local maximum.

Background technology

Under the overall background developing rapidly at intelligent transportation system and technology of Internet of things, realize the robotization that highway and urban road traffic information are processed and controlled, the automated collection systems of transport information is indispensable ingredient in intelligent transportation system, needs a large amount of vehicle equipments to carry out transport information detection.

Vehicle equipment has polytype, comprises video, inductive coil and microwave radar etc.Take linear frequency modulation continuous wave system FMCW(Frequency Modulated Continuous Wave) radar relies on the advantages such as accuracy of detection is high, stability is high, round-the-clock property as the traffic detection technique of sensor, receive increasing concern.Microwave radar, by FMCW principle, gathers radar echo signal, can obtain the range information of detections of radar scope internal object, carries out vehicle discriminating according to the range information of target, obtains the information that exists of vehicle; And then obtain the transport information such as vehicle flowrate, occupation rate and the average speed of surveyed area.

According to linear frequency modulation continuous wave principle, microwave signal is sampled, through Fourier transform, just can obtain each time, be engraved in the reflection wave strength on each parasang.Fig. 2 is the image of the reflection configuration composition that obtains through Fourier transform of many groups fmcw radar signal of actual acquisition.Wherein the horizontal ordinate of image is time shaft, and on time shaft, each column data is the value of one group of Fourier transform, and the echo of being put by different distance forms; Ordinate is distance axis, each in distance a corresponding range unit; The brightness of figure mid point is quantized to calculate by corresponding point radar echo intensity.

Microwave vehicle detecting device is main or according to echo strength to the method for target detection in the market, using the height of echo strength as the criterion that whether has target, distinguishes impact point and non-impact point by setting certain thresholding.But this method makes checkout equipment often occur " false-alarm " and " false dismissal " phenomenon, as the echo traction phenomenon that the cart such as bus, truck is brought, the echo of parallel running vehicle cannot be distinguished phenomenon etc., it is all undesirable that classic method solves effect to these problems, reduced the detection performance of equipment.And due to the otherness of radar antenna, the echo strength of different antennae difference to some extent, therefore adopt former with good grounds intensity carry out vehicle target sentence method for distinguishing can be very high to the coherence request of antenna, this also brings very large difficulty to batch production of product.

Summary of the invention

Goal of the invention: technical matters to be solved by this invention is for the deficiencies in the prior art, provides a kind of Radar for vehicle object detection method based on local maximum.

In order to solve the problems of the technologies described above, the invention discloses a kind of Radar for vehicle object detection method based on local maximum, comprise the following steps:

Microwave detection equipment on step 1, road surface uploads to the echo collecting in host computer by serial ports, observe Ground Penetrating Radar echo distribution situation by host computer, and divide radar echoing area, set affiliated echo region, track, road surface, be saved in the not erasable storer of microwave detection equipment, as later driveway partition foundation setting result; Microwave detection equipment extracts the echo point of track, road surface affiliated area in the echo sequence collecting according to the echo area numeric field data in storer;

Step 2, resolution according to fmcw radar in distance and the width information in track, calculate the scope of search window;

All echo points under the track of extracting in step 3, traversal step 1, search window for each echo point at its neighborhood setting steps 2, then other echo point in this echo point and search window is compared, if the echo strength of this point is maximum, judge that so this point belongs to Local modulus maxima;

Step 4, to the Local modulus maxima searching, judge whether the echo strength value of this point is greater than the echo thresholding of setting, judge that if be greater than this point is impact point, otherwise still think and belong to non-impact point, each range points has an one's own thresholding, the mode that the foundation of thresholding adopts accumulation to average;

Step 5, impact point mark, impact point is labeled as 255, and remaining point, for non-impact point, is labeled as 0, and echoed signal is converted into binary signal, completes the detection of target.

In step 2, calculating the concrete steps of searching plain window ranges is:

Step 2-1, employing FMCW distance-finding method, utilize modulating pulse cycle and modulation voltage amplitude parameter, calculates the resolution L of radar in distance ₁, according to road surface lane width L ₂, calculate echo corresponding to the each track D' that counts out and be: D'=L ₂/ L ₁; Can calculate modulating pulse slope by recurrence interval and modulation voltage according to FMCW principle, according to the difference frequency resolution of antenna and then can calculate range resolution, can list of references: QI G Q.High accuracy range estimation of FMCW level radar based on the phase of the zero-padded FFT[C] .IEEE ICSP04Proceedings, Beijing, 2004:2078-2081.

Step 2-2, search window D is set is greater than the D' calculating in step 2-1, i.e. D=D'+ Δ D, wherein Δ D is 1 or 2 echo points.

The concrete steps that at each echo vertex neighborhood search window are set in step 3 of the present invention are:

Step 3-1, establish the scope of echo point in distance for [0, d _max], wherein 0 correspondence is from the nearest range points of radar detector, d _maxcorresponding to radar detector range points farthest; Establishing its place range points for each echo point is d, and its region of search is [d-Δ d ₁, d+ Δ d ₂], Δ d ₁+ Δ d ₂=D and Δ d ₁than Δ d ₂large 1 or 2;

Step 3-2, hunting zone are in scope [0, d _max] in, meet d-Δ d ₁>=0, and d+ Δ d ₂<=d _max.

The concrete steps of setting echo thresholding in step 4 of the present invention are:

The mode that the foundation of step 4-1, thresholding adopts long time integration to average: for each range points d, accumulate the echo I (d, i) of this point on time shaft, then adopt following mode to calculate the thresholding TH of this point:

$\mathrm{TH}\left(d\right)=K*\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}I(d,i)/n,$

Wherein n represents total scan period number that thresholding is set up, and span is 2000～5000, and each time scan period is 100ms, and corresponding T.T. is 20s～50s, and i represents the current scan period, and K is scale-up factor, and span is 1.5～3.0.

The present invention greatly reduces the often probability of " false-alarm " and " false dismissal " of appearance of fmcw radar vehicle detection process, the echo traction problem of bringing such as the cart such as bus, truck, parallel vehicles travels the echo that brings cannot differentiation problem etc., improves equipment Inspection performance.Make full use of the characteristic distributions of fmcw radar echo in distance simultaneously, reduced the requirement to echo strength in testing process, make detection method there is better adaptability and robustness.

Compared with prior art, its remarkable advantage is this method: (1) takes full advantage of the characteristic distributions of fmcw radar echo in distance axis, extracts the essential characteristic of impact point, has effectively removed the interference that clutter, noise bring.(2) the judgement mode of employing local maximum, can reduce " false-alarm ", such as weeding out in the cart such as public transport, truck motion process traction echo out on other track, even if the echo strength that traction goes out is also more intense, but do not belong to real target point owing to being less than the echo strength of vehicle physical location, still thinking.(3) adopt this method, can reduce " false dismissal ", such as in vehicle parallel running problem, adopt the disposal route of local maximum, can effectively distinguish the middle echo of two cars, two cars is made a distinction, prevent because two cars echo point mixes, it is a car that two cars is detected.(4) having reduced the requirement to echo strength in testing process, is local maximum because become the most important condition of real echo point, but not returns wave intensity.So just make detection algorithm have good adaptability, even if the echo strength of different radar antennas difference is to some extent also very little for the detection impact of vehicle target, improved stability and the robustness of product.

Accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the present invention is done further and illustrated, above-mentioned and/or otherwise advantage of the present invention will become apparent.

Fig. 1 is the process flow diagram that the present invention is based on the vehicle target detection method of local maximum.

Fig. 2 is that image data of the present invention is processed back echo signal two dimension view example.

Fig. 3 is the schematic diagram of employing window maximum searching of the present invention.

Fig. 4 be in the invention process example one bus through out-of-date echo X-Y scheme.

Fig. 5 adopts traditional technique in measuring target simulator result in the invention process example one.

Fig. 6 is the simulation result that adopts the vehicle target detection method based on local maximum that the present invention proposes in the invention process example one.

Fig. 7 be in the invention process example two parallel vehicles simultaneously through out-of-date echo X-Y scheme.

Fig. 8 adopts traditional technique in measuring target simulator result in the invention process example two.

Fig. 9 is the simulation result that adopts the vehicle target detection method based on local maximum that the present invention proposes in the invention process example two.

Figure 10 be in the invention process example three many vehicles through out-of-date echo X-Y scheme.

Figure 11 adopts traditional technique in measuring target simulator result in the invention process example three.

Figure 12 is the simulation result that adopts the vehicle target detection method based on local maximum that the present invention proposes in the invention process example three.

Embodiment

In conjunction with Fig. 1, Fig. 2, a kind of vehicle target detection method based on local maximum the invention provides, comprises the following steps:

The first step, each scan period, microwave signal is sampled, through Fourier transform, just can obtain each time, be engraved in the reflection wave strength on each parasang.Fig. 2 is the image of the reflection configuration composition that obtains through Fourier transform of many groups fmcw radar signal of actual acquisition, can be considered that a coordinate axis is distance, a coordinate axis is the 2D signal of time (scan period), be designated as I (d, i), wherein d is parasang, and i is time (scan period) unit.In actual use, in order to add up information of vehicle flowrate corresponding to each track on road surface, need to calculate or the mode of artificial boundary division is set range points scope corresponding to track on road surface by Algorithm Analysis, concrete operations are as follows: microwave detection equipment uploads to the echo collecting in host computer by serial ports, observe Ground Penetrating Radar echo distribution situation by host computer, and divide radar echoing area, set affiliated echo region, track, road surface, be saved in the not erasable storer of microwave detection equipment setting result, as later driveway partition foundation, microwave detection equipment extracts the echo point of track, road surface affiliated area in the echo sequence collecting according to the echo area numeric field data in storer.

Second step, the resolution according to fmcw radar in distance and the width information in track, the scope of calculating search window.Whether the detection method of the local maximum that this method adopts is in certain window ranges of its neighborhood, to be maximal value according to echo, and then judges whether it is impact point, and " part " is the scope corresponding to window.The scope of window is to determine according to the relativeness between the resolution of radar and lane width, determines the number of the range points that each track occupies.According to fmcw radar parameter, calculate the resolution L of radar in distance ₁, obtain lane width L according to track, road surface situation simultaneously ₂, echo corresponding to each track counted out as D'=L ₂/ L ₁.Consider echo scattering, the situations such as echo traction, search window D should be more bigger than the D' of above-mentioned calculating, D=D'+ Δ D.In practical application, Δ D is generally made as 1 or 2 echo points.

The 3rd step, judges whether echo point belongs to local maximum, and the echo on the different distance point that each scan period is calculated according to the first step adopts the processing mode of sliding window from top to bottom.If the scope of echo point in distance is [0, d in Fig. 2 _max], wherein 0 correspondence the range points nearest from detecting device, and d _maxcorresponding range points farthest.For in each echo point I (d ₀, i), centered by it, on range direction, determine a search window [d ₀-Δ d ₁, d ₀+ Δ d ₂], wherein window size is the window size D that second step calculates, i.e. D=Δ d ₁+ Δ d ₂, schematic diagram as shown in Figure 3.According to field trial result, the phenomenon such as echo scattering and echo traction of target closely goes up even more serious at it in addition, therefore in order better to eliminate the closely impact of echo of impact point, Δ d ₁than Δ d ₂large 1 or 2.Whether the echo strength that then judges this point is the maximal value of echo in window, if this point is designated as maximum point TH _tP(d, i), shown in following formula.

${\mathrm{TH}}_{\mathrm{TP}}(d,i)=\underset{d_0-\mathrm{\&Delta;}{d}_1<d<d_0+\mathrm{\&Delta;}{d}_2}{\mathrm{arc}\max }\left(I(d,i)\right)$

Whether inevitably there will be boundary problem adopting in slide window processing, this method adopts the mode of dwindling window ranges at boundary to process, and prevents that hunting zone from crossing the border, therefore need the echo point that judges index at echo scope [0, d _max] in, should meet d ₀-Δ d ₁>=0, and d ₀+ Δ d ₂<=d _max.

The 4th step, arranges basic thresholding, removes the interference that background dot brings.Because the method for local maximum is maximizing in window, if this window, in background area, even if certain point is maximum value so, but still belongs to background dot, therefore need to arrange a basic thresholding and be used for distinguishing background.Consider that background intensity corresponding to each range points is discrepant, therefore in this method, adopt the method that each range points long time integration echo is averaged to set up basic thresholding.For each range points d, on time shaft, accumulate the echo I (d, i) of this point, then adopt following mode to calculate the thresholding TH of this point:

$\mathrm{TH}\left(d\right)=K*\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}I(d,i)/n$

Wherein n represents total scan period number that thresholding is set up, and span is 2000～5000, and each time scan period is 100ms, and corresponding T.T. is 20s～50s, and i represents the current scan period, and K is scale-up factor, and span is 1.5～3.0.

For each Local modulus maxima, if the basic thresholding that its echo strength is greater than on respective distances point is just judged to be impact point, otherwise be non-impact point.

The 5th step, carries out mark to impact point.Impact point is labeled as 255, and remaining point, for non-impact point, is labeled as 0.So just as follows echoed signal I (d, i) is converted into binary signal I _bIO(d, i), completes the detection of target.

Below in conjunction with Fig. 4 to Figure 12, further illustrate the present invention by embodiment and the effect assessment thereof of three groups of emulation experiments.Data in three groups of embodiment are all the real vehicles echo datas gathering in test site below, and on-the-spot road surface situation is two-way 6 tracks.

Embodiment 1

The reflectogram producing while in the embodiment 1 of Fig. 4 being a bus process detecting device.Wherein horizontal ordinate is time shaft, and each point is corresponding 1 scan period 100ms in time, and this figure was made up of 350 scan periods; Ordinate is distance axis, each in distance corresponding 1 range unit, this figure is made up of 20 range points, each track occupies 3 range points, totally 6 tracks; In figure, the brightness of each point is quantized to calculate by corresponding point radar echo intensity.As can be seen from the figure the echo scope that bus produces distributes very wide, adjacent track a lot of echoes that are also pulled out, and the echo strength being pulled out not a little less than, traditional judge that according to echo strength the method for impact point will detect car on adjacent lane if adopted, and because bus is long through the detecting device time, likely appear at the situation that many cars detected on adjacent lane.

Fig. 5 adopts classic method to carry out the simulation result of target detection, wherein often detects that a car all can stamp a vertical bar mark.5 cars on the adjacent track of bus, detected as we can see from the figure, false-alarm is very high.

Fig. 6 adopts the method based on local maximum of the present invention's proposition to carry out the simulation result of target detection.Could be impact point owing to must meeting the point of local maximum, therefore the echo that traction goes out on adjacent lane is because the echo strength of bus itself is low, still be judged as non-impact point, as can be seen from the figure the method has successfully been eliminated the impact that traction brings, and testing result is correct.

Embodiment 2

The reflectogram producing when to be that two cars is parallel in the embodiment 2 of Fig. 7 cross detecting device simultaneously, this figure is made up of 300 scan periods at time shaft, the same Fig. 4 of distance axis distribution situation.As can be seen from the figure due to two cars from closer, two cars gap cannot obviously be distinguished.

Fig. 8 is the simulation result that adopts traditional technique in measuring, and two cars is judged to be to a car process, occurs false dismissal.

When Fig. 9, adopt the method based on local maximum that the present invention proposes to carry out the simulation result of target detection, can separate vehicle in right area.

Embodiment 3

The reflectogram that the embodiment 3 of Figure 10 produces while being many cars (be three cars in figure, have a car for stopping) process detecting device, this figure is made up of 350 scan periods on time shaft, the same Fig. 4 of distance axis distribution situation.There is sometimes in the process of moving echo scattering phenomenon in vehicle as we can see from the figure, and many cars also there will be echo " multipath " phenomenon together when process.This has increased difficulty to target detection, if adopt traditional detection method based on echo strength will inevitably produce many false-alarms,

Figure 11 is the simulation result of traditional technique in measuring.And the detection method based on local maximum that the present invention proposes, owing to having added the constraint of window extreme value, can eliminate the echo that scattering and " multipath " bring, even if there is a small amount of echo not dispose, in follow-up morphology processing and vehicle discriminating, also can avoid being mistaken for a car.

Figure 12 adopts this method to carry out the simulation result of target detection, and as can be seen from the figure testing result is correct, has effectively eliminated the impact of echo scattering and " multipath ".

Through checking on the spot, the vehicle target detection method based on local maximum that adopts the present invention to propose, the detection correctness of microwave detection equipment under high-speed road conditions reaches 99%, and the detection correctness under the busy road conditions in urban district reaches 95%.

The invention provides a kind of Radar for vehicle object detection method based on local maximum; method and the approach of this technical scheme of specific implementation are a lot; the above is only the preferred embodiment of the present invention; should be understood that; for those skilled in the art; under the premise without departing from the principles of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.In the present embodiment not clear and definite each ingredient all available prior art realized.

Claims (4)

1. the Radar for vehicle object detection method based on local maximum, is characterized in that, comprises the following steps:

Microwave detection equipment on step 1, road surface uploads to the echo collecting in host computer by serial ports, observe Ground Penetrating Radar echo distribution situation by host computer, and divide radar echoing area, set affiliated echo region, track, road surface, be saved in the not erasable storer of microwave detection equipment, as later driveway partition foundation setting result; Microwave detection equipment extracts the echo point of track, road surface affiliated area in the echo sequence collecting according to the echo area numeric field data in storer;

Step 2, resolution according to fmcw radar in distance and the width information in track, calculate the scope of search window;

All echo points under the track of extracting in step 3, traversal step 1, search window for each echo point at its neighborhood setting steps 2, then other echo point in this echo point and search window is compared, if the echo strength of this point is maximum, judge that so this point belongs to Local modulus maxima;

Step 4, to the Local modulus maxima searching, judge whether the echo strength value of this point is greater than the echo thresholding of setting, judge that if be greater than this point is impact point, otherwise still think and belong to non-impact point, each range points has an one's own thresholding, the mode that the foundation of thresholding adopts accumulation to average;

Step 5, impact point mark, impact point is labeled as 255, and remaining point, for non-impact point, is labeled as 0, and echoed signal is converted into binary signal, completes the detection of target.

2. a kind of Radar for vehicle object detection method based on local maximum according to claim 1, is characterized in that, calculates the concrete steps of searching plain window ranges to be in step 2:

Step 2-1, employing FMCW distance-finding method, utilize modulating pulse cycle and modulation voltage amplitude parameter, calculates the resolution L of radar in distance ₁, according to road surface lane width L ₂, calculate echo corresponding to the each track D' that counts out and be: D'=L ₂/ L ₁;

Step 2-2, search window D is set is greater than the D' calculating in step 2-1, i.e. D=D'+ Δ D, wherein Δ D is 1 or 2 echo points.

3. a kind of Radar for vehicle object detection method based on local maximum according to claim 2, is characterized in that, the concrete steps that at each echo vertex neighborhood search window are set in step 3 are:

Step 3-1, establish the scope of echo point in distance for [0, d _max], wherein 0 correspondence is from the nearest range points of radar detector, d _maxcorresponding to radar detector range points farthest; Establishing its place range points for each echo point is d, and its region of search is [d-Δ d ₁, d+ Δ d ₂], Δ d ₁+ Δ d ₂=D and Δ d ₁than Δ d ₂large 1 or 2;

Step 3-2, hunting zone are in scope [0, d _max] in, meet d-Δ d ₁>=0, and d+ Δ d ₂<=d _max.

4. a kind of Radar for vehicle object detection method based on local maximum according to claim 3, is characterized in that, the concrete steps of setting echo thresholding in step 4 are:

The mode that the foundation of step 4-1, thresholding adopts long time integration to average: for each range points d, accumulate the echo I (d, i) of this point on time shaft, then adopt following mode to calculate the thresholding TH of this point:

$\mathrm{TH}\left(d\right)=K*\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}I(d,i)/n,$

Wherein n represents total scan period number that thresholding is set up, and span is 2000～5000, and each time scan period is 100ms, and corresponding T.T. is 20s～50s, and i represents the current scan period, and K is scale-up factor, and span is 1.5～3.0.

Images