Multi-target detection method and system based on vehicle-mounted millimeter-wave radar

Abstract

The present invention relates to the field of target detection technology, and provides a multi-target detection method and system based on a vehicle-mounted millimeter-wave radar. The method comprises: constructing a state transfer function for multi-target detection based on a state model; constructing a likelihood function for multi-target detection based on a measurement model; using the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function. The present invention constructs a particle filter by using a state transfer function and a likelihood function under a random finite set framework, which can effectively avoid interference caused by data association during the filtering process, thereby enabling the multi-target detection task to obtain accurate detection results even in scenarios with clutter and false detection.

Description

Technical Field

[0001] The present invention relates to the field of target detection technology, and in particular to a multi-target detection method and system based on a vehicle-mounted millimeter-wave radar. Background Art

[0002] Autonomous driving, also known as unmanned driving, computer driving or wheeled mobile robot, is a cutting-edge technology that relies on computers and artificial intelligence technology to complete complete, safe and efficient driving without human control.

[0003] In the 21st century, the continued growth of car ownership has led to increasingly severe road congestion and accidents. Autonomous driving, powered by connected vehicles and artificial intelligence, can coordinate travel routes and schedules, significantly improving travel efficiency and reducing energy consumption. Autonomous driving can also help prevent safety hazards such as drunk driving and fatigued driving, reduce driver errors, and enhance safety. Consequently, autonomous driving has become a key research and development priority in recent years.

[0004] As autonomous vehicles, self-driving cars can sense their environment and navigate without human intervention. As a viable form of autonomous driving environmental perception hardware, onboard millimeter-wave radar can collect point cloud data of obstacles during driving. Furthermore, this point cloud data can be used to analyze the state of obstacles, such as the location, speed, and size of multiple targets.

[0005] In autonomous driving or assisted driving scenarios, target association tasks in multi-target detection present certain difficulties and challenges. Existing technologies for target association tasks usually first associate target states and target measurements, and then filter the combinations with the highest association probability.

[0006] However, similar association methods cannot effectively overcome false detection and interference, and are prone to association errors, which can lead to target loss. In other words, the accuracy of target association in existing technologies is difficult to meet the requirements of autonomous driving or assisted driving scenarios, especially high-speed scenarios.

[0007] Therefore, how to provide a multi-target detection method with higher target association accuracy has become a technical problem that needs to be solved urgently in the industry. Summary of the Invention

[0008] The present invention provides a multi-target detection method and system based on a vehicle-mounted millimeter-wave radar, which is used to solve the defects of the prior art such as high false detection and missed detection rates, and realize multi-target detection with higher target association accuracy.

[0009] The present invention provides a multi-target detection method based on a vehicle-mounted millimeter-wave radar, comprising:

[0010] Constructing a state transfer function for multi-target detection based on a state model; the state model is a model for obtaining a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set including state variables of at least two targets; the multi-target predicted state set is a random finite set including predicted values ​​of the state variables of at least two targets;

[0011] Constructing a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that obtains a multi-target measurement set based on a multi-target observation set; the multi-target observation set is a random finite set including observation information collected by an on-board millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set including true information distribution probabilities of at least two targets;

[0012] The multi-target observation set is used as input, and a multi-target detection result is obtained through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function.

[0013] According to a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention, the step of constructing a state transfer function for multi-target detection according to a state model includes:

[0014] Constructing a multi-target detection state transfer function including a first error probability according to the state model;

[0015] The state model is based on an empirical model of uniform linear motion;

[0016] The first error probability is a priori error probability of predicting the motion state transition of multiple targets based on uniform linear motion.

[0017] According to a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention, the step of constructing a likelihood function for multi-target detection based on a measurement model includes:

[0018] Constructing a multi-target detection likelihood function including a second error probability according to the measurement model;

[0019] The measurement model is based on an empirical model of vehicle-mounted millimeter-wave radar parameters;

[0020] The second error probability is a statistical modeling of the error between the observation information collected by the vehicle-mounted millimeter-wave radar and the actual information.

[0021] According to a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention, the step of obtaining a multi-target detection result by a particle filter using the multi-target observation set as input includes:

[0022] Based on the state transfer function, a multi-objective prediction state set at a third moment is obtained;

[0023] Eliminating noise in the multi-objective predicted state set at the third moment based on the drivable area constraint to obtain a constrained particle;

[0024] Obtaining a multi-target measurement set at the third moment according to the multi-target observation set at the third moment and the likelihood function;

[0025] updating the constrained particles according to the multi-objective measurement set to obtain updated particles;

[0026] The updated particles are weighted and averaged, with the probabilities of the updated particles as weights, to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment.

[0027] According to a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention, after the step of weighting and averaging the updated particles using the probabilities of the updated particles as weights to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment, the method further includes:

[0028] resampling the updated particles according to the multi-target detection result at the third moment to obtain resampled particles;

[0029] The distribution density of the resampled particles is proportional to the probability of the updated particles, and the probability of the resampled particles is a set value.

[0030] According to a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention, after the step of weighting and averaging the updated particles using the probabilities of the updated particles as weights to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment, the method further includes:

[0031] If it is determined that the likelihood function has not converged, the likelihood function is updated using a Bayesian recursive formula according to the constrained particles and the multi-target measurement set.

[0032] The present invention also provides a multi-target detection system based on a vehicle-mounted millimeter-wave radar, comprising:

[0033] A state module is configured to construct a state transfer function for multi-target detection based on a state model; the state model is a model for obtaining a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set including state variables of at least two targets; the multi-target predicted state set is a random finite set including predicted values ​​of the state variables of at least two targets;

[0034] A measurement module, configured to construct a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that derives a multi-target measurement set from a multi-target observation set; the multi-target observation set is a random finite set comprising observation information collected by an on-board millimeter-wave radar for at least two targets; and the multi-target measurement set is a random finite set comprising the probability distribution of true information of at least two targets;

[0035] A filtering module is used to obtain a multi-target detection result through a particle filter using the multi-target observation set as input; the particle filter is constructed based on the state transfer function and the likelihood function.

[0036] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, the steps of any of the above-described multi-target detection methods based on vehicle-mounted millimeter-wave radars are implemented.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described multi-target detection methods based on vehicle-mounted millimeter-wave radars.

[0038] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned multi-target detection methods based on vehicle-mounted millimeter-wave radars.

[0039] The multi-target detection method and system based on vehicle-mounted millimeter-wave radar provided by the present invention constructs a particle filter by using a state transfer function and a likelihood function under a random finite set framework, which can effectively avoid the interference caused by data association during the filtering process, thereby enabling the multi-target detection task to obtain accurate detection results in scenarios with clutter and false detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 1 is a flow chart of a multi-target detection method based on a vehicle-mounted millimeter-wave radar provided by the present invention;

[0042] Figure 2 is a schematic diagram of a filtering process provided by an embodiment of the present invention;

[0043] Figure 3 It is a structural diagram of a multi-target detection system based on a vehicle-mounted millimeter-wave radar provided by the present invention;

[0044] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention.

[0045] Reference numerals:

[0046] 1: Status module; 2: Measurement module; 3: Filter module;

[0047] 410: processor; 420: communication interface; 430: memory;

[0048] 440: Communication bus. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0050] The following combination Figure 1 、 Figure 2 The present invention describes a multi-target detection method based on vehicle-borne millimeter-wave radar.

[0051] like Figure 1 As shown, an embodiment of the present invention provides a multi-target detection method based on a vehicle-mounted millimeter-wave radar, comprising:

[0052] Step 101, constructing a state transfer function for multi-target detection based on a state model; the state model is a model that obtains a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set including state variables of at least two targets; the multi-target predicted state set is a random finite set including predicted values ​​of the state variables of at least two targets;

[0053] Step 103: construct a likelihood function for multi-target detection based on the measurement model; the measurement model is a model that obtains a multi-target measurement set based on the multi-target observation set; the multi-target observation set is a random finite set including observation information collected by the vehicle-mounted millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set including the distribution probability of true information of at least two targets;

[0054] Step 105 , using the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function.

[0055] In this embodiment, the execution order of step 103 and step 105 can be adjusted.

[0056] In a preferred embodiment, the first moment is earlier than the second moment, and the first moment and the second moment are adjacent moments related to the acquisition frequency of the vehicle-borne millimeter-wave radar;

[0057] For example, if the acquisition frequency of the vehicle-mounted millimeter-wave radar is 10 Hz, that is, 10 times per second, and the moment of 0 second is taken as the first moment, then the second moment should be the moment of 0.1 second.

[0058] It is worth noting that, in this embodiment, the multi-target state set is the multi-target detection result at that moment. In other words, this embodiment can be understood as a repetitive unit in the iterative process.

[0059] The method of this embodiment can obtain the multi-target detection result at the current moment (i.e., the multi-target state set) through the multi-target detection result at the previous moment (i.e., the multi-target state set) and the multi-target observation set at the current moment.

[0060] If the multi-target detection result (i.e., the multi-target state set) at the previous moment does not exist (for example, when initializing the method of this embodiment, or when multiple targets did not exist at the previous moment), the set initial multi-target state set can be used as the multi-target detection result at the previous moment;

[0061] An optional initial multi-target state set includes initial particles uniformly set within the drivable area, and the probability of each initial particle is a set value (but the sum of the probabilities of the initial particles should be 1).

[0062] In this embodiment, the state transfer function is a mapping of a multi-target state set at a first moment to a multi-target predicted state set at a second moment; wherein, the multi-target state set includes the states of multiple determined targets (such as size, position, speed, etc.); the multi-target predicted state set includes at least one prediction subset, and the prediction subset refers to a set consisting of the states of multiple targets (such as size, position, speed) and the probabilities of multiple targets being in the states, that is, the multi-target predicted state set includes at least one possible state combination and the probability of the state combination.

[0063] Similarly, the likelihood function is a mapping of a multi-target observation set at a set time to a multi-target measurement set at the set time; wherein the multi-target observation set includes the observation values ​​of a determined plurality of targets (such as size, position, velocity, which may also include other parameters that do not belong to the target state, such as Doppler value, RCS value, etc.); the multi-target measurement set includes at least one measurement subset, which refers to a set consisting of the measurement values ​​of a plurality of targets (such as size, position, velocity) and the probability that the true values ​​of the plurality of targets are the same as the measurement values, that is, the multi-target measurement set includes at least one possible measurement value combination and the probability of the measurement value combination.

[0064] Unlike the prior art, in step 105 of this embodiment, the particle filter is a particle filter under a random finite set framework, and each combination of the multi-target prediction state set and the multi-target measurement set can be understood as a possible particle in the particle filtering process. The particle includes a matrix of multi-target states or measurement values, and also includes the probability corresponding to each matrix.

[0065] The beneficial effects of this embodiment are:

[0066] By constructing a particle filter based on the state transfer function and likelihood function under the framework of random finite sets, the interference caused by data association in the filtering process can be effectively avoided, so that the multi-target detection task can still obtain accurate detection results in scenarios with clutter and false detection.

[0067] According to the above embodiment, in this embodiment:

[0068] The step of constructing a state transfer function for multi-target detection according to the state model comprises:

[0069] Constructing a multi-target detection state transfer function including a first error probability according to the state model;

[0070] The state model is based on an empirical model of uniform linear motion;

[0071] The first error probability is a priori error probability of predicting the motion state transition of multiple targets based on uniform linear motion.

[0072] Specifically, the state model is:

[0073]

[0074] Where x and y are the horizontal and vertical coordinates of the target location respectively; v x 、v y are the components of the target velocity on the x and y axes respectively.

[0075] The state transfer function is:

[0076]

[0077] Where, X k+1|k is the target state at the second moment; Δt is the time difference between the two frames; e x 、e y 、e vx 、e vy Represents each state quantity (x, y, v x 、v y ) state error.

[0078] The step of constructing a likelihood function for multi-target detection based on the measurement model includes:

[0079] Constructing a multi-target detection likelihood function including a second error probability according to the measurement model;

[0080] The measurement model is based on an empirical model of vehicle-mounted millimeter-wave radar parameters;

[0081] The second error probability is a statistical modeling of the error between the observation information collected by the vehicle-mounted millimeter-wave radar and the actual information.

[0082] Specifically, the measurement model is:

[0083]

[0084] Where x and y are the horizontal and vertical coordinates of the target location respectively; v x 、v y are the components of the target velocity on the x and y axes respectively

[0085] The likelihood function is:

[0086] fk+1(Z|X)=m!p(m)×S k+1

[0087] Where m is the number of targets, represents the probability of m targets appearing, S k+1 represents the distribution function of the target, which is the joint distribution of each observation of the target; d is the prior expectation of the Poisson distribution; Z is the multi-target measurement set; and e is the base of the natural logarithm.

[0088] The beneficial effects of this embodiment are:

[0089] By introducing the first error probability and the second error probability through the empirical model, the state transfer function and the likelihood function can be made more accurate, thereby improving the accuracy of the subsequent particle filter output value.

[0090] According to any of the above embodiments, in this embodiment:

[0091] like Figure 2 As shown, the step of obtaining a multi-target detection result by a particle filter using the multi-target observation set as input includes:

[0092] Based on the state transfer function, a multi-objective prediction state set at a third moment is obtained;

[0093] Let the third moment be the k-th moment, then the multi-objective prediction state set at the k-th moment satisfies:

[0094]

[0095] Eliminating noise in the multi-objective predicted state set at the third moment based on the drivable area constraint to obtain a constrained particle;

[0096] In other words, particles outside the drivable area in the multi-target predicted state set are clearly erroneous particles. Excluding these clearly erroneous noise particles helps further improve the accuracy of the filter output. In a preferred embodiment, the drivable area is a defined region of interest. For example, the portion of a closed road 100 kilometers away from the vehicle still falls within the broad drivable area. However, in this embodiment, only the narrow drivable area (e.g., the area the vehicle may travel within 10 seconds) that may affect the vehicle's autonomous or assisted driving decisions is considered.

[0097] Obtaining a multi-target measurement set at the third moment according to the multi-target observation set at the third moment and the likelihood function;

[0098] Let the third moment be the k-th moment, and the multi-target measurement set at the k-th moment is Zk.

[0099] The constrained particles are updated according to the multi-target measurement set to obtain updated particles; wherein the particle weight update formula is:

[0100]

[0101] Where fk is the observation likelihood function, is the weight of the i-th particle at time k.

[0102] The updated particles are weighted and averaged, with the probabilities of the updated particles as weights, to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment.

[0103] It is worth noting that the third moment in this embodiment refers to any specific moment, and the third moment cannot be understood as a moment later than the first moment or the second moment, nor can it be understood as a moment other than the first moment or the second moment.

[0104] Furthermore, this embodiment provides a more detailed description of the particle filtering steps as follows.

[0105] At the observation initialization moment, based on the prior information, the state set is initialized for the N0 targets in the area, and the particle set of the state random set at this time is obtained through Gaussian sampling.

[0106] Based on the state transfer function, a predicted state set of the multi-target particle set at the current moment is obtained;

[0107] Eliminating noise in the multi-objective predicted state set at the third moment based on the drivable area constraint to obtain a constrained particle set;

[0108] According to the multi-target observation set at the current moment and the likelihood function, the weight of the particle set is updated to obtain the posterior multi-target particle set at the third moment;

[0109] Integrate the weights of the posterior particle set to obtain the estimated value N of the target number;

[0110] The number of resampling times is determined according to the weight of each particle in the posterior particle set, thereby obtaining the resampled particle set.

[0111] According to the estimated number of targets, the K-means clustering algorithm is used to obtain N target particle sets. The weighted average of the particle states of each particle set is taken as the state value of the N targets to obtain the multi-target detection result at that moment.

[0112] According to the above embodiment, in this embodiment:

[0113] After the step of weighting and averaging the updated particles using the probabilities of the updated particles as weights to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment, the method further includes:

[0114] resampling the updated particles according to the multi-target detection result at the third moment to obtain resampled particles;

[0115] The distribution density of the resampled particles is proportional to the probability of the updated particles, and the probability of the resampled particles is a set value.

[0116] The distribution of resampling particles is set based on the probability of updating particles. The higher the probability of an updating particle, the higher the density of resampling particles within the set range near the location of the updating particle, and vice versa.

[0117] The resampling process may result in an increase in the number of particles, while the constraint process mentioned in the previous embodiment will result in a decrease in the number of particles. If the difference in the number of particles increased or decreased by these two processes is large, it may result in an excessive or insufficient number of particles. Therefore, in a preferred embodiment, if the number of particles is less than the lower threshold, new particles are added using a predetermined method (e.g., uniformly arranged within the drivable area with a predetermined probability) until the total number of particles is no less than the lower threshold; if the number of particles is greater than the upper threshold, particles are deleted using a predetermined method (e.g., deleting particles with a low probability, or randomly deleting particles) until the total number of particles is no more than the upper threshold.

[0118] Furthermore, before the step of weighting and averaging the updated particles using the probability of the updated particles as weights to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment, the method further includes:

[0119] If it is determined that the likelihood function has not converged, the likelihood function is updated using a Bayesian recursive formula according to the constrained particles and the multi-target measurement set.

[0120] If the target detection results are converged, it is confirmed that the target is in a stable tracking state.

[0121] The Bayesian recursion formula is:

[0122] p k|k-1 (X k |Z 1：k-1 )=∫f k|k-1 (X k |ζ)p k-1 (ζ|Z 1：k-1 )μ(dζ)

[0123]

[0124] Among them, f k|k-1 (·|·) and g k (·|·) represent the multi-objective state transfer function and joint likelihood function, respectively, p k|k-1 (·|Z 1：k-1 ) and p k (·|Z 1：k ) represent the multi-target prior and posterior probability density functions respectively.

[0125] Furthermore, since the recursive process of the likelihood function requires integrating the set, the computational complexity of this process will increase sharply as the number of elements in the multi-target measurement set increases. Therefore, there are certain difficulties in the engineering implementation of random set optimal Bayesian filtering.

[0126] For this situation, in a preferred solution of this embodiment, PHD (Probability Hypothesis Density) filtering is used to provide a suboptimal implementation to approximate multi-objective random set Bayesian filtering.

[0127] The multi-target detection device based on a vehicle-mounted millimeter-wave radar provided by the present invention is described below. The multi-target detection device based on a vehicle-mounted millimeter-wave radar described below and the multi-target detection method based on a vehicle-mounted millimeter-wave radar described above can be referenced to each other.

[0128] like Figure 3 As shown, an embodiment of the present invention provides a multi-target detection system based on a vehicle-mounted millimeter-wave radar, comprising:

[0129] State module 1, for constructing a state transfer function for multi-target detection based on a state model; the state model is a model for obtaining a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set including state variables of at least two targets; the multi-target predicted state set is a random finite set including predicted values ​​of the state variables of at least two targets;

[0130] Measurement module 2, configured to construct a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that derives a multi-target measurement set from a multi-target observation set; the multi-target observation set is a random finite set comprising observation information collected by an on-board millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set comprising the probability distribution of true information of at least two targets;

[0131] The filtering module 3 is used to obtain a multi-target detection result through a particle filter using the multi-target observation set as input; the particle filter is constructed based on the state transfer function and the likelihood function.

[0132] Furthermore, the state module 1 includes:

[0133] A state transfer function unit, configured to construct a multi-target detection state transfer function including a first error probability according to the state model;

[0134] The state model is based on an empirical model of uniform linear motion;

[0135] The first error probability is a priori error probability of predicting the motion state transition of multiple targets based on uniform linear motion.

[0136] The measurement module 2 includes:

[0137] A likelihood function unit, configured to construct a multi-target detection likelihood function including a second error probability according to the measurement model;

[0138] The measurement model is based on an empirical model of vehicle-mounted millimeter-wave radar parameters;

[0139] The second error probability is a statistical modeling of the error between the observation information collected by the vehicle-mounted millimeter-wave radar and the actual information.

[0140] The filtering module 3 includes:

[0141] A prediction unit, configured to obtain a multi-objective prediction state set at a third moment based on the state transfer function;

[0142] A constraint unit, configured to eliminate noise in the multi-objective predicted state set at the third moment based on the drivable area constraint to obtain a constrained particle;

[0143] A likelihood unit, configured to obtain a multi-target measurement set at the third moment based on the multi-target observation set at the third moment and the likelihood function;

[0144] a particle updating unit, configured to update the constrained particles according to the multi-target measurement set to obtain updated particles;

[0145] A weighted averaging unit is configured to weightedly average the updated particles using the probabilities of the updated particles as weights, and obtain the multi-target state set at the third moment as the multi-target detection result at the third moment.

[0146] a resampling unit, configured to resample the update particles according to the multi-target detection result at the third moment to obtain resampled particles;

[0147] The distribution density of the resampled particles is proportional to the probability of the updated particles, and the probability of the resampled particles is a set value.

[0148] A recursive unit is configured to determine that the likelihood function has not converged, and then update the likelihood function using a Bayesian recursive formula according to the constrained particles and the multi-target measurement set.

[0149] Figure 4 An example of a physical structure diagram of an electronic device is shown below. Figure 4As shown, the electronic device may include: a processor (processor) 410, a communication interface (Communications Interface) 420, a memory (memory) 430 and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call the logic instructions in the memory 430 to execute a multi-target detection method based on an on-board millimeter-wave radar, which includes: constructing a state transfer function for multi-target detection based on a state model; the state model is a model that obtains a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set, including state variables of at least two targets; the multi-target predicted state set is a random finite set, including predicted values ​​of state variables of at least two targets; constructing a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that obtains a multi-target measurement set based on a multi-target observation set; the multi-target observation set is a random finite set, including observation information collected by the on-board millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set, including the true information distribution probability of at least two targets; taking the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function.

[0150] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0151] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-target detection method based on the vehicle-mounted millimeter-wave radar provided by the above-mentioned methods, the method including: constructing a state transfer function for multi-target detection based on a state model; the state model is a model that obtains a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set that includes state variables of at least two targets; the multi-target predicted state set is a random finite set that includes predicted values ​​of state variables of at least two targets; constructing a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that obtains a multi-target measurement set based on a multi-target observation set; the multi-target observation set is a random finite set that includes observation information collected by the vehicle-mounted millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set that includes the true information distribution probability of at least two targets; using the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function.

[0152] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the multi-target detection method based on the vehicle-mounted millimeter-wave radar provided by the above-mentioned methods, the method comprising: constructing a state transfer function for multi-target detection based on a state model; the state model is a model for obtaining a multi-target predicted state set at a second moment based on a multi-target state set at a first moment; the multi-target state set is a random finite set, including state variables of at least two targets; the multi-target predicted state set is a random finite set, including predicted values ​​of state variables of at least two targets; constructing a likelihood function for multi-target detection based on a measurement model; the measurement model is a model for obtaining a multi-target measurement set based on a multi-target observation set; the multi-target observation set is a random finite set, including observation information collected by the vehicle-mounted millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set, including the true information distribution probability of at least two targets; taking the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function.

[0153] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0154] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A multi-target detection method based on vehicle-mounted millimeter-wave radar, characterized in that: include: Construct the state transfer function of multi-target detection based on the state model; The state model is a model that obtains a multi-objective predicted state set at a second moment based on a multi-objective state set at a first moment; the multi-objective state set is a random finite set including state variables of at least two objectives; the multi-objective predicted state set is a random finite set including predicted values ​​of state variables of at least two objectives; Constructing a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that obtains a multi-target measurement set based on a multi-target observation set; the multi-target observation set is a random finite set including observation information collected by an on-board millimeter-wave radar for at least two targets; the multi-target measurement set is a random finite set including true information distribution probabilities of at least two targets; Taking the multi-target observation set as input, obtaining a multi-target detection result through a particle filter; the particle filter is constructed based on the state transfer function and the likelihood function; Wherein, the state model is: ； Where, 、 are the horizontal and vertical coordinates of the target location respectively; 、 The target speed is 、 The weight of the axis; The state transfer function is: ； Where, is the target state at the second moment; is the time difference between two frames; 、 、 、 Represents each state quantity ( 、 、 、 ) state error.

2. The multi-target detection method based on vehicle-mounted millimeter-wave radar according to claim 1, characterized in that: The step of constructing a state transfer function for multi-target detection according to the state model comprises: Constructing a multi-target detection state transfer function including a first error probability according to the state model; The state model is based on an empirical model of uniform linear motion; The first error probability is a priori error probability of predicting the motion state transition of multiple targets based on uniform linear motion.

3. The multi-target detection method based on vehicle-mounted millimeter-wave radar according to claim 1, characterized in that: The step of constructing a likelihood function for multi-target detection based on the measurement model includes: Constructing a multi-target detection likelihood function including a second error probability according to the measurement model; The measurement model is based on an empirical model of vehicle-mounted millimeter-wave radar parameters; The second error probability is a statistical modeling of the error between the observation information collected by the vehicle-mounted millimeter-wave radar and the actual information.

4. The multi-target detection method based on vehicle-mounted millimeter-wave radar according to claim 1, characterized in that: The step of obtaining a multi-target detection result by a particle filter using the multi-target observation set as input comprises: Based on the state transfer function, a multi-objective prediction state set at a third moment is obtained; Eliminating noise in the multi-objective predicted state set at the third moment based on the drivable area constraint to obtain a constrained particle; Obtaining a multi-target measurement set at the third moment according to the multi-target observation set at the third moment and the likelihood function; updating the constrained particles according to the multi-objective measurement set to obtain updated particles; The updated particles are weighted and averaged, with the probabilities of the updated particles as weights, to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment.

5. The multi-target detection method based on vehicle-mounted millimeter-wave radar according to claim 4, characterized in that: Also includes: resampling the updated particles according to the multi-target detection result at the third moment to obtain resampled particles; The distribution density of the resampled particles is proportional to the probability of the updated particles, and the probability of the resampled particles is a set value.

6. The multi-target detection method based on vehicle-mounted millimeter-wave radar according to claim 4, characterized in that: After the step of weighting and averaging the updated particles using the probabilities of the updated particles as weights to obtain the multi-target state set at the third moment as the multi-target detection result at the third moment, the method further includes: If it is determined that the likelihood function has not converged, the likelihood function is updated using a Bayesian recursive formula according to the constrained particles and the multi-target measurement set.

7. A multi-target detection system based on vehicle-mounted millimeter-wave radar, characterized in that: include: State module, used to construct the state transfer function of multi-target detection based on the state model; The state model is a model that obtains a multi-objective predicted state set at a second moment based on a multi-objective state set at a first moment; the multi-objective state set is a random finite set including state variables of at least two objectives; the multi-objective predicted state set is a random finite set including predicted values ​​of state variables of at least two objectives; A measurement module, configured to construct a likelihood function for multi-target detection based on a measurement model; the measurement model is a model that derives a multi-target measurement set from a multi-target observation set; the multi-target observation set is a random finite set comprising observation information collected by an on-board millimeter-wave radar for at least two targets; and the multi-target measurement set is a random finite set comprising the probability distribution of true information of at least two targets; A filtering module, configured to obtain a multi-target detection result by using a particle filter with the multi-target observation set as input; the particle filter is constructed based on the state transfer function and the likelihood function; Wherein, the state model is: ； Where, 、 are the horizontal and vertical coordinates of the target location respectively; 、 The target speed is 、 The weight of the axis; The state transfer function is: ； Where, is the target state at the second moment; is the time difference between two frames; 、 、 、 Represents each state quantity ( 、 、 、 ) state error.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the multi-target detection method based on vehicle-mounted millimeter-wave radar are implemented as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-target detection method based on vehicle-mounted millimeter-wave radar are implemented as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the multi-target detection method based on vehicle-mounted millimeter-wave radar are implemented as described in any one of claims 1 to 6.

Images