Thermal barrier coatings (TBC) with ceramic topcoat are used to protect jet engine turbine blades and other components against high temperatures, corrosion and erosion wear. Depending on the component and gas turbine type all the coated components are operated in thermal cycling conditions with different parameters. One of the key TBC parameters is thermal cycling durability because state-of-the-art TBC ceramic topcoat material (yttria stabilized zirconia, 7YSZ) has thermal expansion coefficient mismatch with bondcoat material and superalloy substrate. During next generation TBC development coated coupons or components are to be tested by one or several thermal cycling test techniques - furnace, burner rig, laser rig and RF rig are prevailing. Each rig and facility possesses its own features such as heating and cooling speed limits, maximum and minimum operating temperatures, coating condition monitoring etc. The present paper focuses on the computer vision system development aspects for TBC burner rig cycling test application: machine learning process, creating defects database and improving system accuracy. The system developed is now used for coating failure detection.