Incomplete Continuous Rolling Digital Number Recognition Based on Computer Vision

Thereare many problems in the test of modern civil aircraft display system, including many test items, many test versions and large manpower investment. At the same time, often in a short period of time required for a large number of trials, brought test personnel double test of physical and mental. In order to overcome these challenges, improve test efficiency, precision and shorten the test cycle, and reduce the work load of test personnel, for the further research of display system test technology is particularly important. Airspeed information as an important part of instrument symbols in display system, the readings will be as the change of the aircraft speed and scroll up and down, and often can only display part of incomplete rolling numbers, which makes the airspeed information to identify high particularity and challenging work. In view of the display system in modern civil aircraft, airspeed reading take on the characteristics of the incomplete continuous rolling, we proposed a fusion of traditional image recognition with neural network image recognition technology of the new plan. This solution first by image preprocessing step, effectively reduce image noise and improve the accuracy of the subsequent processing. Then, using airspeed digital box positioning technology to accurately determine the coordinates of the four vertices of the digital box. Next, based on these coordinate positions, the image is corrected and the digital area to be recognized is captured. Finally, using the template matching method, the airspeed readings were accurately identified in the captured area, successfully solving the technical problem of incomplete continuous rolling digital image recognition.