CN110146865A - Target identification method and device for radar image - Google Patents
Target identification method and device for radar image Download PDFInfo
- Publication number
- CN110146865A CN110146865A CN201910471034.1A CN201910471034A CN110146865A CN 110146865 A CN110146865 A CN 110146865A CN 201910471034 A CN201910471034 A CN 201910471034A CN 110146865 A CN110146865 A CN 110146865A
- Authority
- CN
- China
- Prior art keywords
- target
- point
- signal
- point cloud
- probability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/414—Discriminating targets with respect to background clutter
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/415—Identification of targets based on measurements of movement associated with the target
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
- Image Analysis (AREA)
Abstract
This specification embodiment provides the target identification method and device for being used for radar image.During carrying out target identification by the signal-tonoise information that signaling point carries for radar image, on the one hand inceptive filtering is carried out to the signaling point in radar image by lower first preset threshold and obtains initial point cloud chart picture, data processing amount can be reduced with preliminary screening, on the other hand, utilize history target identification result, by the historical movement status information of each target, probability distribution point cloud of each target on current radar image is predicted.Further, pass through the probability distribution point cloud of each target, correct the signal-tonoise information of each signaling point in initial point cloud chart picture, and target point cloud chart picture is obtained by screening out the signaling point that signal-to-noise ratio is lower than the level threshold value in revised initial point cloud chart picture, to identify the target in current radar image from target point cloud chart picture.The validity of target identification can be improved in the embodiment.
Description
Technical field
This specification one or more embodiment is related to field of computer technology, more particularly to by computer to radar map
The method and apparatus that target as in is identified.
Background technique
Target identification is either in video, to identify object from one or more image by computer, can also know
Not Chu object state, such as direction, speed.Target identification acquires visual pattern often through camera, or (is swashed by radar
Light, microwave etc.) the means such as point cloud chart picture of reflection acquisition realize.In routine techniques, in the point cloud chart picture acquired by radar
In the case where carrying out target identification, often given threshold, the signal-to-noise ratio for receiving signal to radar are filtered, and reflection is believed
Number signal-to-noise ratio be higher than threshold value signaling point cluster, be identified as the point cloud chart picture of target, and identified, positioned and chased after to target
Track.Signal-to-noise ratio for reflecting signal is lower than the signaling point of threshold value, it is believed that is noise and filters out.
However, in some scenes, if excessively high etc. reasons of object density in detection zone, target to be detected may be by it
His object blocks, and at this moment, even if the penetrability of radar can penetrate object and be reflected by target, however real goal reflects
The signal-to-noise ratio of signal just will be greatly reduced.In routine techniques, if reducing threshold value, Signal with high noise point may be identified as mesh
Mark；If improving threshold value, the corresponding signaling point of real goal is possibly filtered out, causes target can not be by location tracking.
Summary of the invention
This specification one or more embodiment describes a kind of method and apparatus of target identification for radar image,
For solving the problems, such as above-mentioned one or more.
According in a first aspect, providing a kind of method of target identification, which comprises it is current to obtain presumptive area
Radar image, wherein the radar image includes multiple signaling points, and each signaling point carries the letter of corresponding radar appearance
It makes an uproar and compares information；From the multiple signaling point, the signaling point that signal-to-noise ratio is lower than the first preset threshold is screened out, to obtain current
The corresponding initial point cloud chart picture of radar image, wherein first preset threshold is lower than preset level threshold value；For described pre-
Determine each target in region in history target identification result, using corresponding historical movement status information, predicts each target
Each probability distribution point cloud on current radar image, each probability distribution point cloud are described by multiple Probability Points respectively
The probability that each target occurs in corresponding position；According to each probability distribution point cloud to the corresponding letter in the initial point cloud chart picture
The signal-tonoise information of number point is modified；To revised initial point cloud chart picture is passed through, signal-to-noise ratio is screened out lower than the standard threshold
The signaling point of value obtains the target point cloud chart picture for the presumptive area, to identify current thunder from the target point cloud chart picture
Up to the target in image.
In one embodiment, the signal-tonoise information of the corresponding radar appearance of each signaling point passes through corresponding signal point
Gray value indicate, wherein the gray value of each signaling point is directly proportional to corresponding radar appearance.
In one embodiment, each target in the presumptive area in history target identification result includes the first mesh
Mark；It is described to utilize corresponding historical movement status information, predict each probability of each target on current radar image point
Cloud of layouting includes: each movement state information for obtaining the first object in preceding predetermined history target identification result；It will
Each movement state information input prediction model trained in advance, according to the determination of the output result of the prediction model
Probability distribution of the first object at the current radar view acquisition moment.
In one embodiment, the historical movement status information includes at least historical position information, further includes in following
It is one or more: direction, angle, speed.
In one embodiment, the probability that each Probability Point indicates in the probability distribution point cloud is in Gaussian Profile.
It in one embodiment, include multiple targets in history target identification result in the presumptive area；It is described according to
Each probability distribution point cloud in the initial point cloud chart picture corresponding signal point carry out signal-tonoise information be modified include: by
The corresponding each probability distribution point cloud of each target is overlapped, and generates summing point cloud；To in the initial point cloud chart picture
One signaling point, determine in the summing point cloud with consistent first Probability Point of first signal point position, and based on described the
Corresponding first probability of one Probability Point corrects the corresponding signal-tonoise information of first signaling point according to predetermined modification rule.
In one embodiment, the modification rule includes: the case where first probability is less than predetermined probabilities threshold value
Under, reduce mode according to default signal-to-noise ratio and reduces the corresponding signal-to-noise ratio of first signaling point；It is greater than institute in first probability
In the case where stating predetermined probabilities threshold value, increases mode according to default signal-to-noise ratio and increases the corresponding signal-to-noise ratio of first signaling point,
In the default signal-to-noise ratio increase mode, signal-to-noise ratio value added or revised value are positively correlated with first probability.
In one embodiment, the method also includes: to the signaling point in the target point cloud chart picture according to amendment after
Signal-tonoise information and co-ordinate position information clustered；The each cluster that cluster obtains is respectively identified as current radar
Each target in image.
In one embodiment, the method also includes: store the corresponding target identification of current radar image as a result, with
Target identification for subsequent radar image.
According to second aspect, a kind of device of target identification is provided, described device includes:
Acquiring unit is configured to obtain the current radar image of presumptive area, wherein the radar image includes multiple letters
Number point, each signaling point carry the signal-tonoise information of corresponding radar appearance；
First processing units are configured to from the multiple reflection signaling point, screen out signal-to-noise ratio lower than the first preset threshold
Signaling point, to obtain the corresponding initial point cloud chart picture of current radar image, wherein first preset threshold is lower than pre-
If level threshold value；
The second processing unit is configured to for each target in history target identification result in the presumptive area, benefit
With corresponding historical movement status information, each probability distribution point cloud of each target on current radar image is predicted, respectively
A probability distribution point cloud describes the probability that each target occurs in corresponding position by multiple Probability Points respectively；
Amending unit is configured to according to each probability distribution point cloud to the corresponding signal point in the initial point cloud chart picture
Signal-tonoise information is modified；
Recognition unit is configured to screen out signal-to-noise ratio lower than the level threshold value to revised initial point cloud chart picture is passed through
Signaling point obtain the target point cloud chart picture for the presumptive area, to identify current radar from the target point cloud chart picture
Target in image.
According to the third aspect, a kind of computer readable storage medium is provided, computer program is stored thereon with, when described
When computer program executes in a computer, enable computer execute first aspect method.
According to fourth aspect, a kind of calculating equipment, including memory and processor are provided, which is characterized in that described to deposit
It is stored with executable code in reservoir, when the processor executes the executable code, the method for realizing first aspect.
By the target identification method and device of the radar image that this specification embodiment provides, pass through for radar image
During the signal-tonoise information that signaling point carries carries out target identification, on the one hand by lower first preset threshold to radar
Signaling point in image carries out inceptive filtering and obtains initial point cloud chart picture, can reduce data processing amount, another party with preliminary screening
Face, using history target identification as a result, predicting each target in current thunder by the historical movement status information of each target
Up to the probability distribution point cloud on image.Further, it by the probability distribution point cloud of each target, corrects in initial point cloud chart picture
The signal-tonoise information of each signaling point, and it is lower than the level threshold value by screening out signal-to-noise ratio in revised initial point cloud chart picture
Signaling point obtain target point cloud chart picture, to identify the target in current radar image from target point cloud chart picture.In this way, can be with
The noise for preventing the target being blocked from corresponding to signaling point on radar image is relatively low, causes the situation for accidentally deleting erroneous judgement, to mention
The validity of high target identification.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment
Attached drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill of field, without creative efforts, it can also be obtained according to these attached drawings others
Attached drawing.
Fig. 1 shows the implement scene schematic diagram of this specification embodiment；
Fig. 2 shows the processes according to the method for the target identification for radar image of one embodiment；
Fig. 3 shows the corresponding probability distribution point cloud schematic diagram of target when target is human body in a specific example；
Fig. 4 shows the probability distribution schematic diagram of each Probability Point in probability distribution point cloud in a specific example；
Fig. 5 shows the schematic block diagram of the device of the target identification for radar image according to one embodiment.
Specific embodiment
With reference to the accompanying drawing, the scheme provided this specification is described.
Fig. 1 is an implement scene schematic diagram of this specification embodiment.In the implement scene, including presumptive area,
Radar signal acquisition platform, computing platform.Wherein, the radar reflection in presumptive area is acquired by radar signal acquisition platform to believe
Number, and these signals are handled by computing platform, to identify the target in presumptive area, to acquire the corresponding moment
Radar image.Computing platform can be the various hardware devices with certain operational capability or run on the software on hardware device
System etc., such as desktop computer, server etc..It is appreciated that computing platform can also be setting for the compositions such as above equipment
Standby cluster.Computing platform can be set in presumptive area, can also be set to remote service end, it is not limited here.
It is worth noting that radar signal acquisition platform can be any equipment that can acquire radar signal.Itself and meter
Calculating platform can be integrated structure, can also be provided separately.Radar signal acquisition platform for example may include radar transmitter
And receiver, laser radar or microwave (such as millimeter wave) radar are emitted to presumptive area by transmitter, and receive by receiver
Collect the signal that the object in presumptive area returns to radar signal reflections.Here the reflected signal can be known as radar
Reflection signal.At a time, the reflection signal of various objects forms radar image in presumptive area.
Object in presumptive area may disperse to arrange, it is also possible to overlapping arrangement.When overlapped object arrangement, pass through radar
The target that goes out of reflection signal identification may be not comprehensive enough.Therefore, computing platform is when carrying out target identification, first to current radar
Image carries out target identification, by the way that lower threshold is arranged, obtains the rougher initial point cloud chart picture comprising noise, then passes through
The movement state information of respective objects in history point cloud chart picture carries out signal correction to the initial point cloud chart picture at current time, and
Target identification is carried out to revised point cloud chart picture, determines the information such as target and the target state of presumptive area.
In practice, radar image can have certain acquisition time interval, and the time interval is usually and the transmitting category of radar
Property (such as interval of a frame) it is related.Radar emits the electromagnetic wave of certain frequency according to scheduled transmission interval (such as 40 milliseconds), and
The reflection signal by the object in presumptive area is received, radar image is formed.The transmission interval is also possible to radar map simultaneously
The acquisition interval of picture.For the radar image of acquisition of each moment, the offer of this specification embodiment can be executed by computing platform
Target identification method, identify the movement state information of target and target therein.Specifically, if by current radar view
The corresponding acquisition moment is recorded with t, and the initial point cloud chart of t moment is obtained by filtration by coarseness signal for the radar image of t moment
Picture, then computing platform can carry out letter by initial point cloud chart picture of the t-T to the history point cloud chart picture between t moment to t moment
Number amendment.Wherein T is preset time interval, may include the point cloud chart picture of predetermined frame number (such as 5 frames).In some realizations
In, the history point cloud chart picture for carrying out signal correction can also be directly determined according to the predetermined frame number.In this way, at least from
Predetermined frame (may include predetermined frame number before the predetermined frame, such as the 6th frame) point cloud chart picture starts, and each frame point cloud chart picture all may be used
To carry out target identification by adaptive iterative process by the amendment of the point cloud chart picture of preceding predetermined frame number.It is predetermined for the
The point cloud chart picture of predetermined frame number (such as 5 frames) before frame, can be handled, details are not described herein according to predetermined way.This target
The case where knowing otherwise it is possible to prevente effectively from when object blocks target, missing partial target.
One is worth the problem of illustrating to be, the target identification method of this specification embodiment has used historical movement state to believe
Breath, therefore, the target identification method for radar image of this specification is specially adapted to the identification process of moving target.For
Temporarily static target, it is also assumed that the motion state of target is stationary state.
The process of the target identification for radar image is detailed below.
Fig. 2 shows the method flow diagrams according to the target identification of one embodiment.Wherein, Fig. 2 shows method execution
Main body is, for example, the computing platform in Fig. 1.As shown in Fig. 2, the method for target identification is the following steps are included: step 202, obtains pre-
Determine the current radar image in region, wherein radar image includes multiple signaling points, and each signaling point carries corresponding radar reflection
The signal-tonoise information of signal；Step 204, from above-mentioned multiple signaling points, the signal that signal-to-noise ratio is lower than the first preset threshold is screened out
Point, to obtain the corresponding initial point cloud chart picture of current radar image, wherein the first preset threshold is lower than preset standard threshold
Value；Step 206, for each target of history recognition result in presumptive area, using corresponding historical movement status information, in advance
Each probability distribution point cloud of each target on current radar image is surveyed, each probability distribution point cloud passes through multiple general respectively
Rate point describes the probability that each target occurs in corresponding position；Step 208, according to each probability distribution point cloud to initial point cloud
The signal-tonoise information of corresponding signal point in image is modified；Step 210, it to revised initial point cloud chart picture is passed through, sieves
Except signal-to-noise ratio obtains the target point cloud chart picture for presumptive area lower than the signaling point of level threshold value, to know from target point cloud chart picture
Target in not current radar image.
Firstly, in step 202, obtaining the current radar image of presumptive area.Wherein, presumptive area can be radar prison
Control the region in range, such as all areas etc. of a room, radar signal radiation.
It is appreciated that radar can be irradiated target by emitting electromagnetic wave and receive its echo, thus to obtain mesh
It marks to information such as the distance of electromagnetic emission point, range rate (such as radial velocity), orientation, height.Here echo letter
Breath is referred to as reflection signal.When multiple reflection signals flock together, radar image is just depicted.One reflection letter
It number can correspond to a signaling point on radar image.
Signal-to-noise ratio (signal-to-noise ratio, SNR) is the ratio for describing effective component and noise contribution in signal
Relation Parameters, the e.g. power of effective component and noise contribution power ratio.In radar appearance, effective component can be with
It is the signal component reflected through target object, noise contribution can be air atom reflection signal, radar reaches target
Penetrate the interference signal ingredient of the reflection signal of object generation, other signals that such as equipment generates etc.Signal-to-noise ratio can be
Value between 0-1.For signal-to-noise ratio closer to 0, the ingredient of noise signal is bigger, and signal-to-noise ratio closer to 1, get over by the specific gravity of useful signal
Greatly.Therefore, the big signaling point of noise signal ingredient can be filtered out by suitable snr threshold.
In practice, above-mentioned signal-tonoise information can be carried by the signaling point on radar image.For example, signaling point can be passed through
Gray value the signal-to-noise ratio of corresponding radar appearance is described.Gray value can be directly proportional with signal-to-noise ratio.For example, on radar image
The intensity value ranges of signaling point be 0-255, signal-to-noise ratio 0-1 can be respectively corresponded.In one implementation, the determination of gray value
Method is for example are as follows: to 256 × signal-to-noise ratio according to subtracting 1 after round.
Wherein, when Fig. 2 shows process be used for the real-time processing of radar image when, current radar image can be
The radar image at current time.When Fig. 2 shows process be used for the batch processing of history radar image when, current radar map
Picture is also possible to the radar image currently obtained according to chronological order.
Then, by step 204, from above-mentioned multiple signaling points, the signal that signal-to-noise ratio is lower than the first preset threshold is screened out
Point, to obtain the corresponding initial point cloud chart picture of current radar image.It is appreciated that determining target by radar image
Point cloud chart as when, can be screened according to preset snr threshold, herein, by the preset signal-to-noise ratio threshold
Value is known as level threshold value.That is, signal-to-noise ratio is higher than the signaling point of the level threshold value after the level threshold value determines, it is considered
It is the point to form target point cloud, signal-to-noise ratio is considered as noise point lower than the signaling point of the level threshold value.The purpose of this step 204
Be, preliminary screening carried out to the reflection signaling point in radar image, filter out be noise point the bigger signaling point of probability, and
Retain because blocking between object causes noise relatively low it is possible to the signaling point for being mistaken for noise point, it can be with
Select lower snr threshold.That is, the first above-mentioned preset threshold is lower than preset level threshold value.
Wherein, the first preset threshold can also be determined by artificial empirically determined by machine learning method.
By taking machine learning method as an example, can signal-to-noise ratio to the signaling point that history is screened out according to clustering, it is assumed that level threshold value
It for N, can first be clustered respectively centered on M and L, so that the signaling point cluster that history is screened out is 2 classifications, wherein
Cluster classification centered on M is corresponding because of the relatively low signaling point for being mistaken for noise point of noise, the cluster class centered on L
Noise point is not corresponded to.The initial value of M can be N × 3/4, and the initial value of L can be N × 1/4.If with the current value of M
Centered on cluster classification in, be really mistaken for the signaling point of noise point lower than certain proportion (such as 90%), then by M from initial
Value re-starts cluster to the mobile pre- fixed step size (such as 0.01N) of N.Until in cluster classification centered on M, true quilt
The signaling point for being mistaken for noise point reaches above-mentioned certain proportion, takes last time cluster result.If in being with the current value of M
In the cluster classification of the heart, be really mistaken for noise point signaling point be higher than above-mentioned certain proportion, then by M from initial value to
The mobile pre- fixed step size (such as 0.01N) of N/2, and re-start cluster.Until in cluster classification centered on M, it is true misjudged
It is lower than above-mentioned certain proportion for the signaling point of noise point, takes previous cluster result.Then, for the cluster knot taken
Fruit, by the corresponding minimum signal-to-noise ratio of signaling point each in the cluster classification centered on M, as the first preset threshold.
On the other hand, by step 206, for each target in presumptive area in history recognition result, using corresponding
Historical movement status information, predict each probability distribution point cloud of each target on current radar image.Here each
A target can be each target in the target identification result of the radar image of continuous preceding predetermined frame number, before being also possible to
Each target in the target identification result of radar image in predetermined amount of time, is not limited thereto.Each radar image pair
It may include one or more targets in the target identification result answered.
It is appreciated that possibly being present in a target identification result, it is also possible to appear in multiple for each target
In target identification result.Object of which movement has continuity, and for the object continuously moved, the motion state of subsequent time is can be with
According to historical movement state estimations.
Different objects, it is also different to the reflection characteristic of radar.Therefore, it can be carried according to the signaling point in radar image
Different information carry out target identification.Such as it can be by such as to signaling point according to signal-to-noise ratio and the corresponding coordinate of signaling point
It carries out the mode such as clustering and realize.Wherein, the coordinate of the corresponding object reflection point of signaling point usually can be is with radar emission point
The polar coordinates of origin can describe object reflection point at a distance from origin by diameter coordinate, can describe object by angular coordinate
Angle of the reflection point relative to origin.Coordinate information can also be carried in the signaling point of radar image.In this way, every in cluster result
A obtained corresponding object reflection point of class cluster that clusters all is spatially continuous, and can correspond to a target.Here
Target for example can be the various objects of people, vehicle etc..
It is appreciated that the different parts of same target are also different to the reflection characteristic of radar, in some implementations, in order to keep away
Exempt from interfering with each other between target, only the special part of target can also be identified.It, can be with for example, when target is people
By radar setting on higher position (such as roof), it is acquired and knows by the signaling point of the positions reflections such as head to people
Not, to distinguish different people.In such manner, it is possible to which there is deviation (such as in cluster when avoiding partially mutually blocking between human body
The different human bodies being closer gather a class cluster), influence target identification effect.
In radar image, since radar signal has periodicity and continuity, can also be determined in radar image
Motion state of each signaling point, such as current speed, the direction of motion etc..For a target, it may have whole
Motion state.
In one embodiment, the prediction model of historical movement state and current motion state can be trained in advance.Prediction
The training sample of model can be multiple groups target image, and the corresponding target of every group of target image, this group of target image includes pressing
According to the continuously arranged R target image of predetermined time interval, successively by target in the preceding R-1 target image in each group
The selected model of motion state input, and model parameter is adjusted according to the target state in the R target image, to instruct
Practise prediction model.Fig. 2 shows target identification process in, it is assumed that any in presumptive area in history target identification result
Target is first object, each item fortune of the available first object in predetermined (5 such as preceding) history target identification result
These movement state informations are inputted above-mentioned prediction model trained in advance by dynamic status information, and the output result of prediction model is just
It is the estimated location that the first object corresponds to current radar view.These estimated locations can be retouched in a manner of probability distribution
It states.In other words, first object may correspond to each position appeared on current radar image and be retouched by certain probability
It states.Target is referred to as Probability Point in the probability that corresponding position occurs.Multiple Probability Points may be constructed a probability distribution point
Cloud atlas picture.
As an example, referring to FIG. 3, being that a target of a specific example is each general on current radar image
The schematic diagram of rate distributed point cloud.As shown in figure 3, being using human body as in the scene of target, a human body is in estimation sometime
The probability distribution of position.Wherein, shade represents probability, and the most deep part of color is estimated location of the human body at the moment
And motion state, corresponding distribution probability value are bigger.The most shallow part distribution probability of color is close to 0.The probability distribution can be with
For describing centered on target overall estimation position, distribution probability of the target within the scope of predetermined radii.
It in another embodiment, can be according to the continuous characteristics of motion of target, according to the parameter predictions mesh such as direction, speed
Target motion state.For example, according to historical movement state, target is advanced according to constant speed, false in a specific example
If coordinate of some point at the T moment is (r, θ) in target, the point estimation coordinate at prediction T+t moment is (r+VrT, θ+Vθt)。
Wherein, r, θ are the diameter coordinate of the T moment point, angular coordinate, Vr、VθIt is that this is fast radially with circumferential movement at the T moment respectively
Degree, t can be the time interval of radar image acquisition.By the estimated coordinates of the corresponding each point of target, target can be determined
Estimate position and motion state.When determining probability distribution of the target in predetermined position, it is believed that count in the above manner
Each obtained coordinate is the biggish point of target probability of occurrence.
On the probability distribution image of first object, the probability of target appearance can be described by the gray value of each point,
The probability of target appearance can also be described with probability value.When above-mentioned probability distribution is Gaussian Profile, target probability of occurrence is higher
Point concentrate on estimating near position.As shown in figure 4, to concentrate on region 401 corresponding for the biggish Probability Point of target distribution probability
Probability Point, and the probability that Probability Point (such as corresponding Probability Point in the region 402) target of other positions occurs is smaller.At this point, first
Probability summation in the probability distribution image of target in the target appearance of each Probability Point description can also be 1.By with top
Method can determine probability distribution point cloud of each target on current radar view in history target identification result.
Further, in a step 208, it can use the probability distribution point cloud of each target determined in step 206, it is right
The signal-tonoise information of corresponding signal point in initial point cloud chart picture obtained in step 204 is modified.It is appreciated that each mesh
Target probability distribution point cloud depicts the target probability that each location point occurs in current radar view, and target identification mistake
Journey is the signal-to-noise ratio determination based on corresponding signal point in current radar view, therefore, can be based on each probability distribution to initial
The signal-to-noise ratio of corresponding signal point in point cloud chart picture is modified.
It, can be corresponding according to the target in the case where including a target in history target identification result in presumptive area
Probability distribution point cloud, correct the signal-tonoise information of each signaling point in initial point cloud chart picture.The history mesh in presumptive area
It marks in recognition result comprising that can also be superimposed each probability distribution point cloud, with superimposed probability in the case where multiple targets
Distributed point cloud corrects the signal-to-noise ratio that signaling point is accordingly reflected in initial point cloud chart picture.In some implementations, for superimposed general
Rate distributed point cloud, can also be normalized, to guarantee in superimposed probability distribution image, the sum of corresponding probability of each point
It is 1.
In case where including multiple targets in history target identification result in presumptive area, it is assumed that initial point cloud chart picture
In either signal point be the first signaling point, can first determine in summing point cloud with first signal point position consistent first
Probability Point, and it is based on corresponding first probability of the first Probability Point, the corresponding letter of the first signaling point is corrected according to predetermined modification rule
It makes an uproar and compares information.
According to one embodiment, which be can be the corresponding probability of Probability Point each in probability distribution point cloud,
It is converted to the signal-to-noise ratio correction value for accordingly reflecting signaling point in initial point cloud chart picture.Initial point cloud chart corresponding signal point as in
Snr value and the correction value and, so that it may as revised snr value.For example, being multiplied by one by the above probability value
Number, so that the maximum signal to noise ratio in initial point cloud chart picture is no more than 1 with the sum of the most probable value in probability distribution image,
The probability value of each point can serve as the signal-to-noise ratio correction value of corresponding reflection signaling point to the product of the coefficient.
In another embodiment, which may include: the case where the first probability is less than predetermined probabilities threshold value
Under, reduce mode according to default signal-to-noise ratio and reduces the corresponding signal-to-noise ratio of the first signaling point；It is greater than predetermined probabilities threshold in the first probability
In the case where value, increases mode according to default signal-to-noise ratio and increase the corresponding signal-to-noise ratio of the first signaling point.Wherein, in default signal-to-noise ratio
In increase mode, signal-to-noise ratio value added or revised value are positively correlated with the first probability, in default signal-to-noise ratio reduction mode, letter
It makes an uproar more negatively correlated than decreasing value or revised value and the first probability.
With in above-mentioned probability distribution point cloud, the corresponding probability of each Probability Point is in for Gaussian Profile, and target is occurring at this time
The biggish Probability Point of probability compares concentration.As shown in figure 4, probability value at this time, which is distributed between region 401 and region 402, to be had
Obvious drop.If it is determined that the probability value corresponding with the line of demarcation in region 402 of region 401 (predetermined probabilities threshold value), to region
The corresponding Probability Point of each probability value in 401, increases the signal-to-noise ratio of corresponding signal point, and for each probability value in region 402
Corresponding Probability Point reduces the signal-to-noise ratio of corresponding signal point.
In this way, target probability of occurrence can be made greater than the signal-to-noise ratio of the corresponding signaling point of Probability Point of predetermined probabilities threshold value
Value, further increases, and target probability of occurrence is less than the snr value of the corresponding signaling point of Probability Point of predetermined probabilities threshold value, into
One step reduces, to widen gap, reduces high noisy point.
It is worth noting that since probability distribution point cloud is estimated according to target histories motion profile, because
The signaling point in point and initial point cloud chart picture in this probability distribution point cloud is not necessarily stringent correspondence.Further, in determination
When the corresponding relationship of the reflection signaling point in Probability Point and initial point cloud chart picture, it is not necessarily accurate point and point is one-to-one
Relationship.
In one embodiment, region division first can be carried out to the point in probability distribution point cloud, corresponding same probability
Point is divided into the same area.By taking some region marked off as an example, the respective area of initial point cloud chart picture is corresponded to according to the region
Domain, and using the signaling point in corresponding region as the signaling point consistent with the Probability Point in the region.It can be according to the region
The corresponding probability of interior Probability Point is modified the signal-tonoise information of these signaling points.
In another embodiment, each signaling point in initial point cloud chart picture can be mapped to probability distribution point cloud
On image.By taking some signaling point as an example, behind the first position for being mapped to probability distribution point cloud, on probability distribution point cloud chart picture
It is determining with the first position apart from nearest Probability Point, which is determined as to the point consistent with the signaling point, and according to
The corresponding probability of the Probability Point is modified the signal-tonoise information of the signaling point.
In more embodiments, the signaling point in Probability Point and initial point cloud chart picture in probability distribution point cloud is being determined
When corresponding relationship, it can also carry out by other means, details are not described herein.
Then, in step 210, to revised initial point cloud chart picture is passed through, the letter that signal-to-noise ratio is lower than level threshold value is screened out
Number point obtains the target point cloud chart picture for presumptive area, to identify the mesh in current radar image from target point cloud chart picture
Mark.Wherein, level threshold value can be empirically determined, can also be determined according to machine learning model.It, can by taking machine learning as an example
Using by the signaling point in multiple radar images comprising target, as sample, wherein the corresponding signaling point of target is positive sample,
His signaling point (noise point) is negative sample, is carried out according to snr value comprising two clusters by modes such as clusters to positive negative sample
The cluster of classification.Two cluster classifications respectively correspond positive sample classification and negative sample classification.True positive and negative sample is incited somebody to action in cluster result
It, can be by the minimum noise in the corresponding cluster classification of positive sample in the case that this is distinguished according to certain accuracy rate (such as 95%)
Ratio is as level threshold value.
It in some embodiments, can also be to screening out due to may be comprising one or more targets in initial point cloud chart picture
Signal-to-noise ratio is lower than the point cloud chart after the reflection signaling point of level threshold value as further progress target identification.Target identification process can be with
It is carried out by clustering method.Specifically, can to the signaling point in target point cloud chart picture according to revised signal-tonoise information with
And co-ordinate position information is clustered, and each of cluster result cluster is identified as to a mesh in current radar image
Mark.
In a further embodiment, the corresponding target identification of current radar image can also be stored as a result, to be used for
The target identification of subsequent radar image.
Above procedure is looked back, carries out the process of target identification by the signal-tonoise information that signaling point carries for radar image
In, inceptive filtering is on the one hand carried out to the signaling point in radar image by lower first preset threshold and obtains initial point cloud chart
Picture can reduce data processing amount, on the other hand, using history target identification as a result, going through by each target with preliminary screening
History movement state information predicts probability distribution point cloud of each target on current radar image.Further, by each
The probability distribution point cloud of target, corrects the signal-tonoise information of each signaling point in initial point cloud chart picture, and by screening out amendment after
Initial point cloud chart picture in signal-to-noise ratio lower than the signaling point of the level threshold value obtain target point cloud chart picture, with from target point cloud chart
As identifying the target in current radar image.In this way, can prevent the target being blocked from corresponding to signaling point on radar image
Noise it is relatively low, cause the situation for accidentally deleting erroneous judgement, to improve the validity of target identification.
According to the embodiment of another aspect, a kind of Target Identification Unit for radar image is also provided.Fig. 5 shows basis
The schematic block diagram of the Target Identification Unit for radar image of one embodiment.As shown in figure 5, being used for the mesh of radar image
Marking identification device 500 includes:
Acquiring unit 51 is configured to obtain the current radar image of presumptive area, wherein radar image includes multiple signals
Point, each signaling point carry the signal-tonoise information of corresponding radar appearance；
First processing units 52 are configured to from multiple reflection signaling points, screen out signal-to-noise ratio lower than the first preset threshold
Signaling point, to obtain the corresponding initial point cloud chart picture of current radar image, wherein the first preset threshold is lower than preset mark
Quasi- threshold value；
The second processing unit 53 is configured to utilize for each target in history target identification result in presumptive area
Corresponding historical movement status information, predicts each probability distribution point cloud of each target on current radar image, each
Probability distribution point cloud describes the probability that each target occurs in corresponding position by multiple Probability Points respectively；
Amending unit 54 is configured to the letter according to each probability distribution point cloud to the corresponding signal point in initial point cloud chart picture
It makes an uproar and is modified than information；
Recognition unit 55 is configured to screen out signal-to-noise ratio lower than level threshold value to revised initial point cloud chart picture is passed through
Signaling point obtains the target point cloud chart picture for presumptive area, to identify the mesh in current radar image from target point cloud chart picture
Mark.
In one embodiment, the signal-tonoise information of the corresponding radar appearance of each signaling point passes through corresponding signal point
Gray value indicate, wherein the gray value of each signaling point is directly proportional to corresponding radar appearance.
Assuming that each target in presumptive area in history recognition result includes first object, the second processing unit 53 may be used also
To be configured that
Obtain each movement state information of the first object in preceding predetermined history target identification result；
By each movement state information input prediction model trained in advance, the is determined according to the output result of prediction model
Probability distribution of one target at the current radar view acquisition moment.
According to one embodiment, it further includes in following that above-mentioned historical movement status information, which includes at least historical position information,
It is one or more: direction, angle, speed.
In one embodiment, the probability that each Probability Point indicates in the probability distribution point cloud is in Gaussian Profile.
In the case where including multiple targets in history target identification result in presumptive area, amending unit 54 can also match
It is set to:
The corresponding each probability distribution point cloud of each target is overlapped, summing point cloud is generated；
To the first signaling point in initial point cloud chart picture, determine in summing point cloud with the first signal point position consistent first
Probability Point, and it is based on corresponding first probability of the first Probability Point, the corresponding letter of the first signaling point is corrected according to predetermined modification rule
It makes an uproar and compares information.
In a further embodiment, above-mentioned modification rule may include:
In the case where the first probability is less than predetermined probabilities threshold value, reduces mode according to default signal-to-noise ratio and reduce the first signal
The corresponding signal-to-noise ratio of point；
In the case where the first probability is greater than predetermined probabilities threshold value, increases mode according to default signal-to-noise ratio and increase by the first signal
The corresponding signal-to-noise ratio of point, presets in signal-to-noise ratio increase mode, and signal-to-noise ratio value added or revised value are positively correlated with the first probability.
According to one embodiment, recognition unit 55 is also configured as:
Signaling point in target point cloud chart picture is gathered according to revised signal-tonoise information and co-ordinate position information
Class；
By each each target for clustering each obtained cluster and being respectively identified as in current radar image.
According to a possible design, device 500 can also include updating unit, be configured that
The corresponding target identification of current radar image is stored as a result, with the target identification for subsequent radar image.
It is worth noting that device 500 shown in fig. 5 be with Fig. 2 shows the corresponding device of embodiment of the method implement
Example, Fig. 2 shows embodiment of the method in it is corresponding describe be equally applicable to device 500, details are not described herein.
According to the embodiment of another aspect, a kind of computer readable storage medium is also provided, is stored thereon with computer journey
Sequence enables computer execute method described in conjunction with Figure 2 when the computer program executes in a computer.
According to the embodiment of another further aspect, a kind of calculating equipment, including memory and processor, the memory are also provided
In be stored with executable code, when the processor executes the executable code, realize the method in conjunction with described in Fig. 2.
Those skilled in the art are it will be appreciated that in said one or multiple examples, function described in the invention
It can be realized with hardware, software, firmware or their any combination.It when implemented in software, can be by these functions
Storage in computer-readable medium or as on computer-readable medium one or more instructions or code transmitted.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not intended to limit the present invention the foregoing is merely a specific embodiment of the invention
Protection scope, all any modification, equivalent substitution, improvement and etc. on the basis of technical solution of the present invention, done should all
Including within protection scope of the present invention.
Claims (20)
1. a kind of target identification method for radar image, which comprises
Obtain the current radar image of presumptive area, wherein the radar image includes multiple signaling points, and each signaling point carries
The signal-tonoise information of corresponding radar appearance；
From the multiple signaling point, the signaling point that signal-to-noise ratio is lower than the first preset threshold is screened out, to obtain current radar
The corresponding initial point cloud chart picture of image, wherein first preset threshold is lower than preset level threshold value；
For each target in history recognition result in the presumptive area, using corresponding historical movement status information, in advance
Each probability distribution point cloud of each target on current radar image is surveyed, each probability distribution point cloud passes through multiple general respectively
Rate point describes the probability that each target occurs in corresponding position；
It is modified according to signal-tonoise information of each probability distribution point cloud to the corresponding signal point in the initial point cloud chart picture；
To revised initial point cloud chart picture is passed through, screens out signal-to-noise ratio and obtain lower than the signaling point of the level threshold value for described
The target point cloud chart picture of presumptive area, to identify the target in current radar image from the target point cloud chart picture.
2. according to the method described in claim 1, wherein, the signal-tonoise information of the corresponding radar appearance of each signaling point is logical
The gray value for crossing corresponding signal point indicates, wherein the gray value of each signaling point is directly proportional to corresponding radar appearance.
3. according to the method described in claim 1, wherein, each target in the presumptive area includes first object；
It is described to utilize corresponding historical movement status information, predict each probability of each target on current radar image point
Cloud of layouting includes:
Obtain each movement state information of the first object in preceding predetermined history target identification result；
It is true according to the output result of the prediction model by each movement state information input prediction model trained in advance
Probability distribution of the fixed first object at the current radar view acquisition moment.
4. method according to claim 1 or 3, wherein the historical movement status information includes at least history bit confidence
Breath, further includes one or more in following: direction, angle, speed.
5. according to the method described in claim 1, wherein, the probability that each Probability Point indicates in the probability distribution point cloud is in height
This distribution.
6. according to the method described in claim 1, including multiple mesh in history target identification result in the presumptive area wherein
Mark；
It is described according to each probability distribution point cloud in the initial point cloud chart picture corresponding signal point carry out signal-tonoise information into
Row is corrected
The corresponding each probability distribution point cloud of each target is overlapped, summing point cloud is generated；
To the first signaling point in the initial point cloud chart picture, determine in the summing point cloud with first signal point position one
The first Probability Point caused, and it is based on corresponding first probability of first Probability Point, according to predetermined modification rule amendment described the
The corresponding signal-tonoise information of one signaling point.
7. according to the method described in claim 6, wherein, the modification rule includes:
In the case where first probability is less than predetermined probabilities threshold value, reduces mode according to default signal-to-noise ratio and reduce described first
The corresponding signal-to-noise ratio of signaling point；
In the case where first probability is greater than the predetermined probabilities threshold value, increase according to default signal-to-noise ratio increase mode described
The corresponding signal-to-noise ratio of first signaling point, in the default signal-to-noise ratio increase mode, signal-to-noise ratio value added or revised value and institute
State the positive correlation of the first probability.
8. according to the method described in claim 1, wherein, the method also includes:
Signaling point in the target point cloud chart picture is gathered according to revised signal-tonoise information and co-ordinate position information
Class；
The each target each cluster that cluster obtains being respectively identified as in current radar image.
9. according to the method described in claim 1, wherein, the method also includes:
The corresponding target identification of current radar image is stored as a result, with the target identification for subsequent radar image.
10. a kind of Target Identification Unit for radar image, described device include:
Acquiring unit is configured to obtain the current radar image of presumptive area, wherein the radar image includes multiple signals
Point, each signaling point carry the signal-tonoise information of corresponding radar appearance；
First processing units are configured to from the multiple reflection signaling point, screen out the letter that signal-to-noise ratio is lower than the first preset threshold
Number point, to obtain the corresponding initial point cloud chart picture of current radar image, wherein first preset threshold is lower than preset
Level threshold value；
The second processing unit is configured to utilize phase for each target in history target identification result in the presumptive area
The historical movement status information answered predicts each probability distribution point cloud of each target on current radar image, each general
Rate distributed point cloud describes the probability that each target occurs in corresponding position by multiple Probability Points respectively；
Amending unit is configured to the noise according to each probability distribution point cloud to the corresponding signal point in the initial point cloud chart picture
It is modified than information；
Recognition unit is configured to screen out the letter that signal-to-noise ratio is lower than the level threshold value to revised initial point cloud chart picture is passed through
Number point obtains the target point cloud chart picture for the presumptive area, to identify current radar image from the target point cloud chart picture
In target.
11. device according to claim 10, wherein the signal-tonoise information of the corresponding radar appearance of each signaling point
It is indicated by the gray value of corresponding signal point, wherein the gray value of each signaling point is directly proportional to corresponding radar appearance.
12. device according to claim 10, wherein each target in the presumptive area includes first object；
Described the second processing unit is further configured to:
Obtain each movement state information of the first object in preceding predetermined history target identification result；
It is true according to the output result of the prediction model by each movement state information input prediction model trained in advance
Probability distribution of the fixed first object at the current radar view acquisition moment.
13. device described in 0 or 12 according to claim 1, wherein the historical movement status information includes at least historical position
Information further includes one or more in following: direction, angle, speed.
14. device according to claim 10, wherein the probability of each Probability Point expression is in the probability distribution point cloud
Gaussian Profile.
15. device according to claim 10, wherein comprising multiple in history target identification result in the presumptive area
Target；
The amending unit is further configured to:
The corresponding each probability distribution point cloud of each target is overlapped, summing point cloud is generated；
To the first signaling point in the initial point cloud chart picture, determine in the summing point cloud with first signal point position one
The first Probability Point caused, and it is based on corresponding first probability of first Probability Point, according to predetermined modification rule amendment described the
The corresponding signal-tonoise information of one signaling point.
16. device according to claim 15, wherein the modification rule includes:
In the case where first probability is less than predetermined probabilities threshold value, reduces mode according to default signal-to-noise ratio and reduce described first
The corresponding signal-to-noise ratio of signaling point；
In the case where first probability is greater than the predetermined probabilities threshold value, increase according to default signal-to-noise ratio increase mode described
The corresponding signal-to-noise ratio of first signaling point, in the default signal-to-noise ratio increase mode, signal-to-noise ratio value added or revised value and institute
State the positive correlation of the first probability.
17. device according to claim 10, wherein the recognition unit is additionally configured to:
Signaling point in the target point cloud chart picture is gathered according to revised signal-tonoise information and co-ordinate position information
Class；
The each target each cluster that cluster obtains being respectively identified as in current radar image.
18. device according to claim 10, wherein described device further includes updating unit:
The corresponding target identification of current radar image is stored as a result, with the target identification for subsequent radar image.
19. a kind of computer readable storage medium, is stored thereon with computer program, when the computer program in a computer
When execution, computer perform claim is enabled to require the method for any one of 1-9.
20. a kind of calculating equipment, including memory and processor, which is characterized in that be stored with executable generation in the memory
Code realizes method of any of claims 1-9 when the processor executes the executable code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910471034.1A CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910471034.1A CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110146865A true CN110146865A (en) | 2019-08-20 |
| CN110146865B CN110146865B (en) | 2023-07-14 |
Family
ID=67590178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910471034.1A Active CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110146865B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091148A (en) * | 2019-12-10 | 2020-05-01 | 成都纳雷科技有限公司 | Vehicle-mounted radar target tracking method and device based on target prior information |
| CN111158014A (en) * | 2019-12-30 | 2020-05-15 | 华通科技有限公司 | Multi-radar comprehensive bird detection system |
| CN111781608A (en) * | 2020-07-03 | 2020-10-16 | 浙江光珀智能科技有限公司 | Moving target detection method and system based on FMCW laser radar |
| CN112285665A (en) * | 2020-10-14 | 2021-01-29 | 广州极飞科技有限公司 | Signal detection method and device, millimeter wave radar module, equipment and medium |
| CN113657331A (en) * | 2021-08-23 | 2021-11-16 | 深圳科卫机器人科技有限公司 | Warning line infrared induction identification method and device, computer equipment and storage medium |
| CN113793307A (en) * | 2021-08-23 | 2021-12-14 | 上海派影医疗科技有限公司 | Automatic labeling method and system suitable for multi-type pathological images |
| CN113920778A (en) * | 2021-12-15 | 2022-01-11 | 深圳佑驾创新科技有限公司 | Image acquisition method and device |
| CN115410408A (en) * | 2022-08-31 | 2022-11-29 | 南京慧尔视智能科技有限公司 | Method, device, equipment and medium for detecting parking space state change |
| WO2023108544A1 (en) * | 2021-12-15 | 2023-06-22 | 深圳航天科技创新研究院 | Single-antenna ultra-wideband radar system for imaging application |
| CN116626630A (en) * | 2023-07-25 | 2023-08-22 | 北京赛目科技股份有限公司 | Object classification method and device, electronic equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101173987A (en) * | 2007-10-31 | 2008-05-07 | 北京航空航天大学 | Multi-module and multi-target accurate tracking apparatus and method thereof |
| CN106054169A (en) * | 2016-05-18 | 2016-10-26 | 西安电子科技大学 | Multi-station radar signal fusion detection method based on tracking information |
| WO2017143217A1 (en) * | 2016-02-18 | 2017-08-24 | Aeye, Inc. | Adaptive ladar receiver |
| US20170254896A1 (en) * | 2016-03-02 | 2017-09-07 | Oki Electric Industry Co., Ltd. | Tracking apparatus, tracking method, and computer-readable storage medium |
| CN108152831A (en) * | 2017-12-06 | 2018-06-12 | 中国农业大学 | A kind of laser radar obstacle recognition method and system |
| WO2018105842A1 (en) * | 2016-12-09 | 2018-06-14 | 메타빌드(주) | Radar-based high-precision incident detection system |
| CN108230379A (en) * | 2017-12-29 | 2018-06-29 | 百度在线网络技术（北京）有限公司 | For merging the method and apparatus of point cloud data |
| CN108469609A (en) * | 2018-06-11 | 2018-08-31 | 成都纳雷科技有限公司 | A kind of detection information filtering method for radar target tracking |
| CN108509918A (en) * | 2018-04-03 | 2018-09-07 | 中国人民解放军国防科技大学 | Target detection and tracking method fusing laser point cloud and image |
| CN109100741A (en) * | 2018-06-11 | 2018-12-28 | 长安大学 | A kind of object detection method based on 3D laser radar and image data |
| CN109190469A (en) * | 2018-07-27 | 2019-01-11 | 阿里巴巴集团控股有限公司 | A kind of detection method and device, a kind of calculating equipment and storage medium |
-
2019
- 2019-05-31 CN CN201910471034.1A patent/CN110146865B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101173987A (en) * | 2007-10-31 | 2008-05-07 | 北京航空航天大学 | Multi-module and multi-target accurate tracking apparatus and method thereof |
| WO2017143217A1 (en) * | 2016-02-18 | 2017-08-24 | Aeye, Inc. | Adaptive ladar receiver |
| US20170254896A1 (en) * | 2016-03-02 | 2017-09-07 | Oki Electric Industry Co., Ltd. | Tracking apparatus, tracking method, and computer-readable storage medium |
| CN106054169A (en) * | 2016-05-18 | 2016-10-26 | 西安电子科技大学 | Multi-station radar signal fusion detection method based on tracking information |
| WO2018105842A1 (en) * | 2016-12-09 | 2018-06-14 | 메타빌드(주) | Radar-based high-precision incident detection system |
| CN108152831A (en) * | 2017-12-06 | 2018-06-12 | 中国农业大学 | A kind of laser radar obstacle recognition method and system |
| CN108230379A (en) * | 2017-12-29 | 2018-06-29 | 百度在线网络技术（北京）有限公司 | For merging the method and apparatus of point cloud data |
| CN108509918A (en) * | 2018-04-03 | 2018-09-07 | 中国人民解放军国防科技大学 | Target detection and tracking method fusing laser point cloud and image |
| CN108469609A (en) * | 2018-06-11 | 2018-08-31 | 成都纳雷科技有限公司 | A kind of detection information filtering method for radar target tracking |
| CN109100741A (en) * | 2018-06-11 | 2018-12-28 | 长安大学 | A kind of object detection method based on 3D laser radar and image data |
| CN109190469A (en) * | 2018-07-27 | 2019-01-11 | 阿里巴巴集团控股有限公司 | A kind of detection method and device, a kind of calculating equipment and storage medium |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091148B (en) * | 2019-12-10 | 2021-07-27 | 成都纳雷科技有限公司 | Vehicle-mounted radar target tracking method and device based on target prior information |
| CN111091148A (en) * | 2019-12-10 | 2020-05-01 | 成都纳雷科技有限公司 | Vehicle-mounted radar target tracking method and device based on target prior information |
| CN111158014A (en) * | 2019-12-30 | 2020-05-15 | 华通科技有限公司 | Multi-radar comprehensive bird detection system |
| CN111158014B (en) * | 2019-12-30 | 2023-06-30 | 华通科技有限公司 | Multi-radar comprehensive bird detection system |
| CN111781608A (en) * | 2020-07-03 | 2020-10-16 | 浙江光珀智能科技有限公司 | Moving target detection method and system based on FMCW laser radar |
| CN111781608B (en) * | 2020-07-03 | 2023-04-25 | 浙江光珀智能科技有限公司 | Moving target detection method and system based on FMCW laser radar |
| CN112285665B (en) * | 2020-10-14 | 2023-12-26 | 广州极飞科技股份有限公司 | Signal detection method and device, millimeter wave radar module, equipment and medium |
| CN112285665A (en) * | 2020-10-14 | 2021-01-29 | 广州极飞科技有限公司 | Signal detection method and device, millimeter wave radar module, equipment and medium |
| CN113657331A (en) * | 2021-08-23 | 2021-11-16 | 深圳科卫机器人科技有限公司 | Warning line infrared induction identification method and device, computer equipment and storage medium |
| CN113793307A (en) * | 2021-08-23 | 2021-12-14 | 上海派影医疗科技有限公司 | Automatic labeling method and system suitable for multi-type pathological images |
| CN113920778A (en) * | 2021-12-15 | 2022-01-11 | 深圳佑驾创新科技有限公司 | Image acquisition method and device |
| WO2023108544A1 (en) * | 2021-12-15 | 2023-06-22 | 深圳航天科技创新研究院 | Single-antenna ultra-wideband radar system for imaging application |
| CN115410408A (en) * | 2022-08-31 | 2022-11-29 | 南京慧尔视智能科技有限公司 | Method, device, equipment and medium for detecting parking space state change |
| CN115410408B (en) * | 2022-08-31 | 2024-05-24 | 南京慧尔视智能科技有限公司 | Parking space state change detection method, device, equipment and medium |
| CN116626630B (en) * | 2023-07-25 | 2023-09-29 | 北京赛目科技股份有限公司 | Object classification method and device, electronic equipment and storage medium |
| CN116626630A (en) * | 2023-07-25 | 2023-08-22 | 北京赛目科技股份有限公司 | Object classification method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110146865B (en) | 2023-07-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110146865A (en) | Target identification method and device for radar image | |
| US7990311B2 (en) | Adaptive clutter filter for maritime surface search radar | |
| NL2012327B1 (en) | Selective intrusion detection systems. | |
| US8502731B2 (en) | System and method for moving target detection | |
| CN110992298A (en) | Genetic algorithm-based radiation source target identification and information analysis method | |
| JP2017156219A (en) | Tracking device, tracking method, and program | |
| US10032285B1 (en) | Multi-hypothesis moving object detection system | |
| JP2014002085A (en) | Signal processing device, radar device, signal processing method, and program | |
| CN110400294B (en) | Infrared target detection system and detection method | |
| CN111624567A (en) | Constant false alarm detection method and device | |
| KR102340184B1 (en) | Apparatus and method for tracking a manual target using AI learning technique in the console terminal | |
| Ai et al. | A new two parameter CFAR ship detector in Log-Normal clutter | |
| US10281573B1 (en) | Retrodiction tracking system and related techniques | |
| JP2012103197A (en) | Moving target adaptive type scan correlation method | |
| CN115712306A (en) | Unmanned aerial vehicle navigation method for multi-machine cooperation target tracking | |
| KR20150109882A (en) | Method and apparatus for detecting object in ultra wide band radar | |
| CN111123251A (en) | Target object detection method and device of radar | |
| CN111157954B (en) | Self-adaptive adjustment method for ground radar target detection threshold coefficient | |
| US20220373672A1 (en) | Multipath classification in radar detections | |
| US10788583B2 (en) | Rapid robust detection decreaser | |
| JP2021015079A (en) | Radar monitoring device and method | |
| US20220383146A1 (en) | Method and Device for Training a Machine Learning Algorithm | |
| JP2019197051A (en) | Target, method for tracking target especially for tracking maritime target, and radar for implementing such method | |
| CN111480180A (en) | Method for detecting and tracking objects | |
| CN111665499A (en) | Judgment method for reducing marine oil spill early warning misjudgment rate of shipborne radar |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |