Approaches for Sensor Data Fusion in Thermal Imaging to Improve the Measurement Quality

Dr.-Ing. Samuel Soldan

April 2010 - August 2014

BMWi and Land Hessen

Thermography has established itself as an important tool for inspection, testing and monitoring. In many applications just a single measurement/image is often insufficient, for example because structures cannot be perceived clearly enough, if the target is larger than the field of view of the camera optics or because characteristics arise only from the chronological trend in the measurements. In addition to that, based on his experience many other factors are captured by the camera operator with his senses and that are also considered in the decision-making. Automating the inspection tasks in unknown environments is therefore a very complex endeavour.

In this work sensor data fusion approaches involving thermography are analysed and discussed. The goal of these approaches is to improve the quality of the measurement and to increase the amount of usable information about a target. The fusion can be carried out in different ways that can be categorized based on the used sensors and if the sensor is fixed or allowed to change its orientation/position. With measurements only from the thermal imaging camera it is possible for example

• to reduce noise by temporal averaging,
• to increase the depth of field by combining data with different focus length,
• to increase the temperature measuring range by combining measurements obtained with different temperature range settings or
• to achieve suppression of artefacts/reflections and to derive geometric information and surface properties by changing filter settings.

Further, by changing the orientation of the camera

• the field of view can be enlarged or
• the geometric resolution can be improved.

Sensor data fusion of thermograms can also be carried out using data from other imaging instruments, for example with

• normal cameras in the visible range,
• thermography cameras in other wavelength ranges or
• depth cameras/rangefinders.

For each measured point more information can be gained and the resulting accumulated data can be interpreted and analysed geometrically as 3D models. The multitude of information can enable a solid documentation and good (automatic) analysis.

Besides the different sensor data fusion approaches the basics of thermography and the registration of measured data are shown. In an application example the feasibility of automated inspection on a mobile robot and the combination and practical application of the various sensor data fusion approaches are shown.