A Point Cloud Enhancement Method for 4D mmWave Radar Imagery

Qingmian Wan,Hongli Peng,Xing Liao,Kuayue Liu,Junfa Mao

2024-01-30

Abstract:A point cloud enhancement method for 4D mmWave radar imagery is proposed in this paper. Based on the patch antenna and MIMO array theories, the MIMO array with small redundancy and high SNR is designed to provide the probability of high angular resolution and detection rate. The antenna array is deployed using a ladder shape in vertical direction to decrease the redundancy and improve the resolution in horizontal direction with the constrains of physical factors. Considering the complicated environment of the real world with non-uniform distributed clutters, the dynamic detection method is used to solve the weak target sensing problem. The window size of CFAR detector is assumed variant to be determined using optimization method, making it adaptive to different environments especially when weak targets exist. The angular resolution increase using FT-based DOA method and the designed antenna array is described, which provides the basis of accurate detection and dense point cloud. To verify the performance of the proposed method, experiments of simulations and practical measurements are carried out, whose results show that the accuracy and the point cloud density are improved with comparison of the original manufacturer mmWave radar of TI AWR2243.

Image and Video Processing

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to solve several key problems in 4D millimeter - wave (mmWave) radar point - cloud images for autonomous driving applications, specifically including:

1. **Improve point - cloud density and accuracy**: The point - cloud data generated by existing 4D millimeter - wave radars has a low density in complex environments, resulting in insufficient accuracy in target detection, recognition, and tracking. The paper proposes a method to enhance the point - cloud to improve its density and accuracy.

2. **Improve weak - target detection ability**: In a non - uniformly distributed clutter environment, the detection of weak targets is a challenge. The paper introduces a dynamic CFAR (Constant False Alarm Rate) detection method, which improves the detection performance of weak targets by optimizing the window size and adaptively adjusting the threshold.

3. **Enhance angular resolution**: In order to achieve more accurate spatial positioning, the paper designs an optimized antenna array, which reduces redundancy and improves angular resolution. This provides a basis for the high - precision generation of subsequent point - clouds.

4. **Optimize antenna design**: The paper proposes a MIMO antenna - array design based on a trapezoidal layout to reduce redundancy and improve the signal - to - noise ratio (SNR), thereby achieving higher angular resolution and detection rates.

### Specific problem descriptions

- **Ambiguity problems in multi - target scenarios**: In multi - target scenarios, the radar faces the problem of separating targets from the environment, especially in complex clutter environments.
- **Weak - target detection**: The existing coherent phase - difference method cannot handle the detection problem of moving weak targets, especially in non - stationary states.
- **Point - cloud sparsity and multi - reflection effects**: In complex environments, it is difficult to distinguish targets due to the sparsity of point - cloud images and multi - reflection effects.

### Solutions

- **Optimize antenna design and array deployment**: By designing a MIMO antenna array with small redundancy and high SNR, the angular resolution and detection performance are improved.
- **Dynamic CFAR detection**: The dynamic CFAR algorithm is introduced, allowing the detection window size to be adaptively adjusted according to the environment, improving the accuracy of weak - target detection.
- **DOA estimation optimization**: By optimizing the antenna array, the DOA angular resolution is improved, thereby achieving more accurate spatial point positioning.
- **Experimental verification**: Through simulation and actual measurement experiments, the improvement effects of the proposed methods in terms of resolution, point - cloud density, and spatial accuracy are verified.

In summary, this paper is mainly dedicated to improving the point - cloud generation quality of 4D millimeter - wave radars in complex environments, especially for the needs of weak - target detection and high - precision spatial positioning.