| 摘要: |
| 针对直升机载毫米波雷达在复杂环境下新老类型障碍物(群目标、孤立目标及电力线)检测与识别困难的问题，提出了一种环境适应性强的目标检测识别算法，通过将电力塔目标与电力线目标检测识别处理相结合来提高电力线的检测概率。该方法首先对雷达图像去噪，提取无噪声图像划分连通域，然后应用恒虚警(Constant False Alarm Rate,CFAR)检测雷达图像得到疑似目标点，并基于图像连通域对疑似目标点进行凝聚。基于凝聚目标提取并计算目标的图像特征，利用离线机器学习得到特征知识库，利用特征知识库完成目标分类，识别出集群目标、电力塔及其他孤立目标。最后根据电力线布拉格散射特性识别电力线目标，并利用霍夫（Hough）直线检测对电力塔及电力线进行确认。该毫米波防撞雷达目标检测识别算法在复杂背景环境下具有良好的适应性，能够对新老类型障碍物完成检测识别，即使是电力线目标淹没在背景环境时，也能利用电力塔目标联合处理完成电力线提取，识别概率达98%。 |
| 关键词: 直升机防撞雷达；毫米波雷达；电力线识别 电力塔识别；机器学习；霍夫直线检测 |
| DOI：10.20079/j.issn.1001-893x.240424001 |
|
| 基金项目: |
|
| A Target Detection and Recognition Algorithm for Helicopter-borne Millimeter-wave Collision Avoidance Radar |
| PENG Xinyi |
| (Southwest China Institute of Electronic Technology，Chengdu 610036，China) |
| Abstract: |
| The millimeter-wave（MMW） radar mounted on helicopters faces challenges in detecting and identifying old and new types of obstacles such as group targets,isolated targets,and power lines in complex environments.To address this issue,a highly adaptive target detection and recognition algorithm is proposed which combines the detection and recognition of power tower targets with power line targets to enhance the detection probability of power lines.Firstly,the radar image is de-noised to extract a noise-free image,and connected regions are identified within it.Subsequently,the constant false alarm rate(CFAR)algorithm is applied to the radar image to detect potential target points,which are then grouped into coherent targets based on the connected regions.Image features of the grouped targets are extracted and calculated.An offline machine learning process generates a feature knowledge base,which is utilized to classify the targets,including cluster targets,power towers and other isolated targets.Finally,power line targets are identified based on the Bragg scatter characteristics,and the power towers and power lines are confirmed using the Hough line detection algorithm.The proposed MMW collision avoidance radar target detection and recognition algorithm exhibits excellent adaptability in complex backgrounds environments.It is capable of detecting and identifying both new and old types of obstacles,and even when power line targets are submerged in the background,it can utilize joint processing of power tower targets to accomplish power line extraction with a recognition probability of 98 |
| Key words: helicopter-borne collision avoidance radar millimeter-wave radar power line recognition power tower recognition machine learning Hough line detection |
