Manipulation of Camera Sensor Data via Fault Injection for Anomaly Detection Studies in Verification and Validation Activities For AI

TL;DR

This paper presents a database of 10,000 images created by injecting seven different fault types into robot camera data, aimed at improving anomaly detection in robotic systems during verification and validation.

Contribution

The study introduces a novel fault-injected image database for robotic cameras, facilitating enhanced anomaly detection methods in robotic system validation.

Findings

Database contains 10,000 images with balanced normal and faulty data.

Faults include erosion, dilation, opening, closing, gradient, motion blur, and partial loss.

Provides a resource for developing and testing anomaly detection algorithms.

Abstract

In this study, the creation of a database consisting of images obtained as a result of deformation in the images recorded by these cameras by injecting faults into the robot camera nodes and alternative uses of this database are explained. The study is based on an existing camera fault injection software that injects faults into the cameras of a working robot and collects the normal and faulty images recorded during this injection. The database obtained in the study is a source for the detection of anomalies that may occur in robotic systems. Within the scope of this study, a database of 10000 images consisting of 5000 normal and 5000 faulty images was created. Faulty images were obtained by injecting seven different types of image faults, namely erosion, dilation, opening, closing, gradient, motionblur and partialloss, at different times while the robot was operating.