We present a machine-learning approach to detect and analyze meteor echoes (MADAME), which is a radar data processing workflow featuring advanced machine-learning techniques using both supervised and unsupervised learning. Our results demonstrate that YOLOv4, a convolutional neural network (CNN)-based one-stage object detection model, performs remarkably well in detecting and identifying meteor head and trail echoes within processed radar signals. The detector can identify more than 80 echoes per minute in the testing data obtained from the Jicamarca high power large aperture (HPLA) radar. MADAME is also capable of autonomously processing data in an interferometer mode, as well as determining the target’s radiant source and vector velocity. In the testing data, the Eta Aquarids meteor shower could be clearly identified from the meteor radiant source distribution analyzed automatically by MADAME, thereby demonstrating the proposed algorithm’s functionality. In addition, MADAME found that about 50 percent of the meteors were traveling in inclined and near-inclined circular orbits. Furthermore, meteor head echoes with a trail are more likely to originate from shower meteor sources. Our results highlight the capability of advanced machine-learning techniques in radar signal processing, providing an efficient and powerful tool to facilitate future and new meteor research.
A novel computer vision‐based meteor head echo detection algorithm is developed to study meteor fluxes and their physical properties, including initial range, range coverage, and radial velocity. The proposed Algorithm for Head Echo Automatic Detection (AHEAD) comprises a feature extraction function and a Convolutional Neural Network (CNN). The former is tailored to identify meteor head echoes, and then a CNN is employed to remove false alarms. In the testing of meteor data collected with the Jicamarca 50 MHz incoherent scatter radar, the new algorithm detects over 180 meteors per minute at dawn, which is 2 to 10 times more sensitive than prior manual or algorithmic approaches, with a false alarm rate less than 1 percent. The present work lays the foundation of developing a fully automatic AI‐meteor package that detects, analyzes, and distinguishes among many types of meteor echoes. Furthermore, although initially evaluated for meteor data collected with the Jicamarca VHF incoherent radar, the new algorithm is generic enough that can be applied to other facilities with minor modifications. The CNN removes up to 98 percent of false alarms according to the testing set. We also present and discuss the physical characteristics of meteors detected with AHEAD, including flux rate, initial range, line of sight velocity, Signal‐to‐Noise Ratio, and noise characteristics. Our results indicate that stronger meteor echoes are detected at a slightly lower altitude and lower radial velocity than other meteors.
more » « less- PAR ID:
- 10377169
- Publisher / Repository:
- DOI PREFIX: 10.1029
- Date Published:
- Journal Name:
- Radio Science
- Volume:
- 57
- Issue:
- 10
- ISSN:
- 0048-6604
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
-
more » « less
Abstract High‐power large‐aperture radar instruments are capable of detecting thousands of meteor head echoes within hours of observation, and manually identifying every head echo is prohibitively time‐consuming. Previous work has demonstrated that convolutional neural networks (CNNs) accurately detect head echoes, but training a CNN requires thousands of head echo examples manually identified at the same facility and with similar experiment parameters. Since pre‐labeled data is often unavailable, a method is developed to simulate head echo observations at any given frequency and pulse code. Real instances of radar clutter, noise, or ionospheric phenomena such as the equatorial electrojet are additively combined with synthetic head echo examples. This enables the CNN to differentiate between head echoes and other phenomena. CNNs are trained using tens of thousands of simulated head echoes at each of three radar facilities, where concurrent meteor observations were performed in October 2019. Each CNN is tested on a subset of actual data containing hundreds of head echoes, and demonstrates greater than 97% classification accuracy at each facility. The CNNs are capable of identifying a comprehensive set of head echoes, with over 70% sensitivity at all three facilities, including when the equatorial electrojet is present. The CNN demonstrates greater sensitivity to head echoes with higher signal strength, but still detects more than half of head echoes with maximum signal strength below 20 dB that would likely be missed during manual detection. These results demonstrate the ability of the synthetic data approach to train a machine learning algorithm to detect head echoes.
-
more » « less
Abstract On 10 and 11 October 2019, high‐power radar observations were performed simultaneously for 8 hours at Resolute Bay Incoherent Scatter North (RISR‐N), Jicamarca Radio Observatory (JRO), and Millstone Hill Observatory (MHO). The concurrent observations eliminate diurnal, seasonal, and space weather biases in the meteor head echo populations and elucidate relative sensitivities of each facility and configuration. Each facility observed thousands of head echoes, with JRO observing tens of thousands. An inter‐pulse phase matching technique employs Doppler shifts to determine head echo range rates (velocity component along radar beam) with order‐of‐magnitude greater accuracy versus measuring the Doppler shift at individual pulses, and this technique yields accurate range rates and decelerations for a subset of the head echo population at each facility. Because RISR‐N is at high latitude and points away from the ecliptic plane, it does not observe head echoes with range rates faster than 55 km/s, although its head echo population demonstrates a bias toward larger and faster head echoes. At JRO near the equator, a larger spread of range rates is observed. MHO observes a large spread of range rates at mid‐latitude despite its comparable frequency to RISR‐N, but this occurs because its beam was pointed at a 45° elevation angle unlike RISR‐N and JRO which were pointed near‐zenith. A trend of greater decelerations at lower altitudes is observed at RISR‐N and JRO, with decelerations of up to 60 km/s2, but high‐deceleration events of up to 1,000 km/s2previously observed in head echo studies are not observed.
-
more » « less
ABSTRACT This work presents the result of sporadic meteor radiant density distribution using the Arecibo 430 MHz incoherent scatter radar (ISR) located in Puerto Rico for the first time. Although numerous meteor studies have been carried out using the Arecibo ISR, meteoroid radiant density distribution has remained a mystery as the Arecibo radar cannot measure vector velocity. A numerical orbital simulation algorithm using dynamic programming and stochastic gradient descent is designed to solve the sporadic meteoroid radiant density and the corresponding speed distributions of the meteors observed at Arecibo. The data set for the algorithm comprises over 250 000 meteors from Arecibo observations between 2009 and 2017. Five of the six recognized sporadic meteor sources can be identified from our result. There is no clearly identifiable South Apex source. Instead, there is a broad distribution between +/−30° ecliptic latitude, with the peak density located in the North Apex direction. Our results also indicate that the Arecibo radar is not sensitive to meteors travelling straight into or perpendicular to the antenna beam but is most sensitive to meteors with an arrival angle between 30° and 60°. Our analysis indicates that about 75 per cent of meteoroids observed by the Arecibo radar travel in prograde orbits when the impact probability is considered. Most of the retrograde meteoroids travel in inclined low-eccentricity orbits.
-
more » « less
Abstract We investigate the mechanism underlying lower hybrid waves associated with high altitude echoes recently detected in the post‐sunset equatorial topside ionosphere and inner plasmasphere by the Jicamarca VHF radar. These waves are visible as prominent sidebands in the echo Doppler spectra. New experimental results and newly processed incoherent scatter radar (ISR) datasets are presented that provide clues as to the conditions in which the echoes and associated waves occur. Numerical simulations are presented which demonstrate the feasibility of an inverse energy cascade coupled with a short wavelength instability, that is, the lower hybrid drift instability, in explaining the waves. An inverse cascade is required for short wavelength lower hybrid waves to extend to the 3 m wavelengths measured by the Jicamarca radar. The simulations were able to reproduce some features of the measurements including the lower hybrid sidebands at 3 m wavelengths, asymmetry in the sidebands, and the damping effect of higher densities and lower altitudes.
