Art Conservation Meets Engineering

Problem

Art conservators assess and restore defects on hundreds of paintings every year at museums and art galleries. Defect analysis is conducted when artworks travel between galleries every year for exhibitions. The National Gallery of Canada estimates that a total of 800 to 1000 artworks are exchanged via a loan program every year. During the transit phase, damage can occur across any layer of the painting, with cracking being the most common deteriorations. There is a need for a faster, easy to use and accurate mechanism for detection of cracks in paintings.

Analysis Time

Current analysis times range from 30 min to 4 hours per art piece. Goal is to minimize both survey and analysis times.

Crack detection

Current analysis techniques are capable of detecting cracks visible to the human eye. At the very least crack widths of 10μm should be detected.

Costs

iPads are currently primarily used by conservators to survey and analyse a painting for defects. The total product costs must therefore be approximately equivalent to the cost of an iPad Pro.

Current analysis methods are time intensive and require well experienced analysts!

Solution

This project aims to deliver a solution for automating the survey and analysis process of defect detection in paintings, saving a museum thousands of hours in analysis time a year. Automatic crack detection is the specific goal of this project, to be performed with the usage of consumer grade DSLR mounted on a gantry system. The capture of the images and detection of defects is conducted in real time.

Core Components

Software: Crack Detection

Key function is for the detection of cracks of minimum width 10μm. Processing of the images captured by the camera is conducted on a device (ex. laptop) separate from the scanning apparatus. A defect map of all existing cracks should be produced for the entire painting for the conservator to view and assess.
Sensors

Key function is to capture a set of images of the painting with 10 um details using the camera and flash ring of LED lights. These images are then offloaded to a server where they are stored and available for processing in real time or in a later period by the detection software.
Mechanical structure

Key functions include securing the sensor, removal of ambient light, securing the painting and vibration isolation within the system and from the environment. Support the movement of the camera along the major and minor axes.
Motors and Power Management

Key functions include provision of minimum major and minor axes resolution of ±1 mm by the motors and provision of two major voltage rails, 24V and 5V, for the motors, linear actuators, Raspberry Pi and Arduino.