This article summary is inspired from this article 'How CAMS, the Cameras for Allsky Meteor Surveillance Project, detects long-period comets through machine learning' and paper 'SpaceML: Distributed Open-source Research with Citizen Scientists for the Advancement of Space Technology for NASA'

Please Don't Forget To Join Our 1600+ ML Subreddit

Long-period comets (LPCs) are comets with orbital periods greater than 200 years. According to scientists, they are difficult to discover and pose a significant threat to the Earth’s ecosystem. Therefore, it has attracted a lot of interest from scientists who aim to provide early notification of a likely impact.

The Cameras for All-sky Meteor Surveillance (CAMS) project started online in October 2010 at Fremont Peak Observatory in California with the installation of low-light camera systems. For this, images were recorded from these cameras using a specially developed compression system. They then went through a software utility designed to detect the presence of meteors. This allowed the researchers to spot evidence of long-period comets that other observing approaches might miss.

Despite the use of software, the monitoring procedure required significant human intervention, with data only being collected from the sites every two months. Daily updates, in which observations from one night are made available to the scientific community the next day for review, would require a different approach: machine learning.

The existing AI pipeline accessible to CAMS sites is designed to reduce the amount of work a human operator has to perform.

New search for SpaceML, an extension of NASA’s Frontier Development Lab artificial intelligence accelerator, now introduces a new six-stage pipeline. The researchers used machine learning and deep learning technologies to improve and automate the classification of meteors from non-meteors. Their goal was to remove the human factor from the CAMS data processing pipeline while maintaining the accuracy of the processed results. The researchers explain the steps as follows:

1. First stage: Local devices at operator locations that capture sky data undertake local processing to assess whether a reported object is a meteor or non-meteor. They also process these, including clouds, planes, and birds, which could cause false detection in the system. They achieved precision and recall ratings of ninety and ninety percent. The team used the following:

  • A random forest classifier with binary meteor or non-meteor classification
  • Convolutional neural network that produces a probability score for a series of image frames
  • A long-term memory network (LSTM) is designed to predict the probability of lightcurve traces corresponding to a meteor.

2. Step Two – Data Recovery: The data is then retrieved from the remote site, which previously required bi-monthly in-person visits and retrieval of the physical DVD media to which the data was burned.

3. Step Three – Treatment: It is executed by Python scripts that interact with and automate the existing CAMS software stack, including MeteorCal, installed cameras and star observations.

4. Fourth step – Calculation of coincidence: This process takes confirmed meteors and combines data from multiple cameras to create a trajectory, automatically recognizing and correcting irregularities in the video recording that could lead to inaccuracies. The automated approach aims to reduce the amount of human labor involved in the process by using classifiers that examine slight curve shapes and maximum errors in geographic positions.

5. Step Five – Data Aggregation: The aim is to identify bursts and new showers that could suggest the presence of a hitherto unknown long-period comet. The pipeline can detect previously unidentified meteor shower clusters and potential meteor explosions using the t-Stochastic Neighbor Embedding (t-SNE) approach for unsupervised machine learning to process parameters. This is followed by spatial clustering based on the density of applications with noise (DBSCAN) for identification of clusters.

6. Step Six – Visualization Step: CAMS data that has gone through the previous five steps is transformed into a more accessible form to make it more widely available. This last step takes the data and places it in a freely rotating sphere. This sphere can be viewed using custom-written JavaScript in any current web browser, including smartphones and tablets.

This ability to compare current activity to historical data makes it easy to spot unexpected behavior and new rains. The proposed approach expanded its scope to include data from observing stations with as few as one or two cameras and process it through the same automated pipeline, allowing the project to expand its sky monitoring coverage.