Automated Detection of Lunar Rockfalls Using a Convolutional Neural Network

This paper implements a novel approach to automatically detect and classify rockfalls in Lunar Reconnaissance Orbiter narrow angle camera (NAC) images using a single-stage dense object detector (RetinaNet). The convolutional neural network has been trained with a data set of 2932 original rockfall images. In order to avoid overfitting, the initial training data set has been augmented during training using random image rotation, scaling, and flipping. Testing images have been labelled by human operators and have been used for RetinaNet performance evaluation. Testing shows that RetinaNet is capable to reach recall values between 0.98 and 0.39, precision values between 1 and 0.25, and average precisions ranging from 0.89 to 0.69, depending on the used confidence threshold and intersection-over-union values. Mean processing time of a single NAC image in RetinaNet is around 10 s using a GeForce GTX 1080 Ti and GeForce Titan Xp, which is in orders of magnitudes faster than a human operator. The processing speed allows to efficiently exploit the currently available NAC data stack with more than 1 million images in a reasonable timeframe. The combination of speed and detection performance can be used to produce lunar rockfall distribution maps on large spatial scales for utilization by the scientific and engineering community.

I. Introduction

Lunar rockfalls and their traces have been investigated by researchers since the Lunar Orbiter Program sent back the first high-resolution images of the lunar surface in 1966 [1]–[7]. In addition to images taken from orbital platforms, astronauts Eugene Cernan and Harrison Schmitt were able to directly sample and photograph boulder and track remnants of rockfalls in the Taurus–Littrow valley during the Apollo 17 mission in 1972 [8]. Since the arrival of NASA’s Lunar Reconnaissance Orbiter (LRO) in orbit around the moon in 2009 [9], a growing repository of high-resolution images taken by the LRO’s narrow angle camera (NAC) is available that enables the studies of geomorphological processes and small-scale features such as rockfalls in unprecedented detail. Various investigations use NAC imagery to identify and study lunar rockfalls as a tool to, e.g., estimate local seismic activity and shaking intensity in Schrödinger Crater [10] or to investigate the topographic evolution of the moon [11]. Besides that these studies show that rockfalls are a widespread phenomenon and occur in various geologic settings.