| dc.description.abstract | In the process of software development, in order to repair, refactor or add functions, engineers will constantly modify the existing code. To ensure the quality of the software after every change, regression testing is very important. so as to allow testers to perform regression testing more efficiently, the laboratory has developed a set of recording/playback testing tool - Korat.
For achieve non-platform dependence, Korat uses image recognition to detect targets. On non-Windows platforms, we use the image capture card to obtain the screen of the system under test as the image source for image recognition. We found that tools such as Korat, which are based on computer vision to identify target, have GUIs in specific situations, such as anti-aliasing, target image changes, or image capture card signal noise interference, resulting in the current test image and recording test image is different, the test case cannot be played back normally.
In this thesis, we will explain the specific situation of the test case that cannot be correctly executed for recording/playback test tool Korat, and propose an automatic learning and correction script method to automatically increase the target picture and use the method of image recognition result set. Reduce the impact of the above problems, and finally explain the direction that can be improved in the future. | en_US |