A method and system based on point cloud data simulation between different millimeter-wave radars

Abstract

The present invention relates to the field of intelligent driving technology, and specifically discloses a method based on point cloud data simulation between different millimeter-wave radars, including: respectively obtaining historical point cloud data of a first millimeter-wave radar and a second millimeter-wave radar, and inputting them into an autoencoder model for training; judging whether the currently trained model reaches the set simulation accuracy; if so, obtaining model parameters to obtain a trained autoencoder model, and then inputting the current point cloud data of the first millimeter-wave radar into the trained model to obtain simulated point cloud data of the second millimeter-wave radar; otherwise, obtaining more historical point cloud data to continue training the current model until the set simulation accuracy is reached. The present invention also discloses a system based on point cloud data simulation between different millimeter-wave radars. The method based on point cloud data simulation between different millimeter-wave radars provided by the present invention can achieve the purpose of mutual conversion of image signal point clouds.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and more specifically, to a method for simulating point cloud data between different millimeter-wave radars and a system for simulating point cloud data between different millimeter-wave radars. Background Art

[0002] With the continuous development of science and technology and people's constant pursuit of living standards, more and more research is being conducted in the field of intelligent driving. Intelligent driving is mainly divided into perception system, decision-making system and control system. The perception system mainly uses sensors to perceive and detect moving targets (vehicles, non-motorized vehicles, pedestrians) and stationary targets (telephone poles, traffic lights, road concrete curbs, iron fences, elevated sound insulation boards) in the road environment to obtain information in the environment.

[0003] Currently, the mainstream automotive millimeter-wave radar frequency bands in China and abroad are 24 GHz (for short- and medium-range radar, 15-30 meters) and 77 GHz (for long-range radar, 100-200 meters). This application has become widespread in European, American, and Japanese vehicles, with nearly all vehicles in these regions now equipped with automotive millimeter-wave radar sensors, including those for collision avoidance and blind-spot detection. These sensors vary in price. However, the point cloud signals output by these various radars are also quite different. Therefore, how to simulate the point cloud signals of expensive radars with low-cost radars, or in some cases, integrate the functions of low-cost radars with high-cost radars, has become a very interesting research topic. Summary of the Invention

[0004] In order to address the deficiencies in the prior art, the present invention provides a method and system for simulating point cloud data between different millimeter-wave radars. By using deep learning to establish a mapping relationship between two millimeter-wave radars of different models, the purpose of mutual conversion of image signal point clouds can be achieved.

[0005] As a first aspect of the present invention, a method for simulating point cloud data between different millimeter-wave radars is provided, comprising:

[0006] Step S1: Acquire historical point cloud data of the first millimeter-wave radar;

[0007] Step S2: Acquire historical point cloud data of the second millimeter-wave radar;

[0008] Step S3: selecting historical point cloud data within the first millimeter-wave radar setting space and storing it as a first high-dimensional vector, and then inputting the first high-dimensional vector into the AutoEncoder model for training;

[0009] Step S4: selecting historical point cloud data within the second millimeter-wave radar setting space and storing it as a second high-dimensional vector, and then inputting the second high-dimensional vector into the AutoEncoder model for training;

[0010] Step S5: Starting the training of the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector;

[0011] Step S6: Determine whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, execute step S8; otherwise, execute step S7;

[0012] Step S7: obtaining more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy;

[0013] Step S8: Obtain the trained AutoEncoder model parameters to obtain the trained AutoEncoder model;

[0014] Step S9: inputting the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar;

[0015] Step S10: End.

[0016] Furthermore, the first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

[0017] As a second aspect of the present invention, a system based on point cloud data simulation between different millimeter-wave radars is provided, comprising:

[0018] A first acquisition module is used to respectively acquire historical point cloud data of the first millimeter-wave radar and historical point cloud data of the second millimeter-wave radar;

[0019] a selection module, configured to select historical point cloud data within the setting space of the first millimeter-wave radar, save the data as a first high-dimensional vector, and then input the first high-dimensional vector into the AutoEncoder model for training; and simultaneously select historical point cloud data within the setting space of the second millimeter-wave radar, save the data as a second high-dimensional vector, and then input the second high-dimensional vector into the AutoEncoder model for training;

[0020] A training module, configured to start training the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector;

[0021] A judgment module is used to judge whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, the execution action of the third acquisition module is performed; otherwise, the execution action of the second acquisition module is performed;

[0022] A second acquisition module is used to acquire more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy;

[0023] The third acquisition module is used to obtain the trained AutoEncoder model parameters and obtain the trained AutoEncoder model;

[0024] A simulation module is used to input the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar.

[0025] Furthermore, the first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

[0026] The method and system for simulating point cloud data between different millimeter-wave radars provided by the present invention have the following advantages:

[0027] 1. Using different 4D millimeter-wave radars to map each other improves the detection of target properties;

[0028] 2. By mapping different 4D millimeter-wave radars with each other, low-cost radars can be used to generate the point cloud effect of high-cost radars, reducing the cost of radar purchase;

[0029] 3. By mapping different 4D millimeter-wave radars with each other, the amount of data can be reduced when there are a large number of point clouds, the data throughput can be reduced, and communication efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0031] Figure 1 The figure is a flow chart of the method for simulating point cloud data between different millimeter-wave radars based on the present invention. DETAILED DESCRIPTION

[0032] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of the method and system for simulating point cloud data between different millimeter-wave radars proposed in accordance with the present invention. Obviously, the described embodiments are only a portion of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0033] In this embodiment, a method based on point cloud data simulation between different millimeter wave radars is provided. Figure 1 As shown, the method based on point cloud data simulation between different millimeter-wave radars includes:

[0034] Step S1: Acquire historical point cloud data of the first millimeter-wave radar;

[0035] Step S2: Acquire historical point cloud data of the second millimeter-wave radar;

[0036] Step S3: selecting historical point cloud data within the first millimeter-wave radar setting space and storing it as a first high-dimensional vector, and then inputting the first high-dimensional vector into the AutoEncoder model for training;

[0037] It should be noted that after obtaining the historical point cloud data of the first millimeter wave radar, the length, width and height are set to select the historical point cloud data within a certain length, width and height space;

[0038] Step S4: selecting historical point cloud data within the second millimeter-wave radar setting space and storing it as a second high-dimensional vector, and then inputting the second high-dimensional vector into the AutoEncoder model for training;

[0039] It should be noted that after obtaining the historical point cloud data of the second millimeter-wave radar, the same length, width and height are set as those of the first millimeter-wave radar, and the historical point cloud data in the same length, width and height space are selected;

[0040] Step S5: Starting the training of the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector;

[0041] Step S6: Determine whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, execute step S8; otherwise, execute step S7;

[0042] Step S7: obtaining more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy;

[0043] Step S8: Obtain the trained AutoEncoder model parameters to obtain the trained AutoEncoder model;

[0044] Step S9: inputting the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar;

[0045] Step S10: End.

[0046] It should be noted that the point cloud data generated by the first millimeter-wave radar is denoted as matrix A*. The matrix A* is compared element by element with the point cloud data matrix B* generated by the second millimeter-wave radar, and a fluctuation factor is introduced. The number of qualified elements is divided by the total number of elements. If the final result exceeds the set threshold, it is considered that the current AutoEncoder model has achieved the set point cloud data simulation accuracy.

[0047] Preferably, the first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

[0048] It should be understood that the AutoEncoder model is an automatic encoder model.

[0049] As another embodiment of the present invention, a system based on point cloud data simulation between different millimeter-wave radars is provided, which includes:

[0050] A first acquisition module is used to respectively acquire historical point cloud data of the first millimeter-wave radar and historical point cloud data of the second millimeter-wave radar;

[0051] a selection module, configured to select historical point cloud data within the setting space of the first millimeter-wave radar, save the data as a first high-dimensional vector, and then input the first high-dimensional vector into the AutoEncoder model for training; and simultaneously select historical point cloud data within the setting space of the second millimeter-wave radar, save the data as a second high-dimensional vector, and then input the second high-dimensional vector into the AutoEncoder model for training;

[0052] A training module, configured to start training the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector;

[0053] A judgment module is used to judge whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, the execution action of the third acquisition module is performed; otherwise, the execution action of the second acquisition module is performed;

[0054] A second acquisition module is used to acquire more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy;

[0055] The third acquisition module is used to obtain the trained AutoEncoder model parameters and obtain the trained AutoEncoder model;

[0056] A simulation module is used to input the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar.

[0057] Preferably, the first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

[0058] The present invention provides a method for simulating point cloud data between different millimeter-wave radars, and in particular relates to a deep learning simulation algorithm based on the mutual mapping of different 4D millimeter-wave radars. The method can use the target points generated by millimeter-wave radar A to imitate the target points generated by millimeter-wave radar B, thereby reducing the difference in point cloud quality between different 4D millimeter-wave radars.

[0059] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment, it is not intended to limit the present invention. Any technician familiar with the present profession can make slight changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method based on point cloud data simulation between different millimeter-wave radars, characterized in that: include: Step S1: Acquire historical point cloud data of the first millimeter-wave radar; Step S2: Acquire historical point cloud data of the second millimeter-wave radar; Step S3: selecting historical point cloud data within the first millimeter-wave radar setting space and storing it as a first high-dimensional vector, and then inputting the first high-dimensional vector into the AutoEncoder model for training; Step S4: selecting historical point cloud data within the second millimeter-wave radar setting space and storing it as a second high-dimensional vector, and then inputting the second high-dimensional vector into the AutoEncoder model for training; Step S5: Starting the training of the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector; Step S6: Determine whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, execute step S8; otherwise, execute step S7; Step S7: obtaining more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy; Step S8: Obtain the trained AutoEncoder model parameters to obtain the trained AutoEncoder model; Step S9: inputting the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar; Step S10: End.

2. The method for simulating point cloud data between different millimeter-wave radars according to claim 1, characterized in that: The first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

3. A system based on point cloud data simulation between different millimeter-wave radars, characterized in that: include: A first acquisition module is used to respectively acquire historical point cloud data of the first millimeter-wave radar and historical point cloud data of the second millimeter-wave radar; a selection module, configured to select historical point cloud data within the setting space of the first millimeter-wave radar, save the data as a first high-dimensional vector, and then input the first high-dimensional vector into the AutoEncoder model for training; and simultaneously select historical point cloud data within the setting space of the second millimeter-wave radar, save the data as a second high-dimensional vector, and then input the second high-dimensional vector into the AutoEncoder model for training; A training module, configured to start training the AutoEncoder model based on the first high-dimensional vector and the second high-dimensional vector; A judgment module is used to judge whether the currently trained AutoEncoder model has reached the set point cloud data simulation accuracy. If so, the execution action of the third acquisition module is performed; otherwise, the execution action of the second acquisition module is performed; A second acquisition module is used to acquire more first high-dimensional vectors and second high-dimensional vectors to continue training the current AutoEncoder model until the current AutoEncoder model reaches the set point cloud data simulation accuracy; The third acquisition module is used to obtain the trained AutoEncoder model parameters and obtain the trained AutoEncoder model; A simulation module is used to input the current point cloud data of the first millimeter-wave radar collected in real time into the trained AutoEncoder model to obtain simulated point cloud data of the second millimeter-wave radar.

4. The system based on point cloud data simulation between different millimeter-wave radars according to claim 3, characterized in that: The first millimeter-wave radar and the second millimeter-wave radar both include 4D millimeter-wave radars.

Images