A system to detect color changes using sequential image processing

A system to detect step-like color changes in sequences of video image data was developed and tested. The system is directed toward real-time application to monitor liquid fueled rocket engines such as the space shuttle main engine. The system consists of a highpass filter cascaded with a moving-average filter applied to each of the three color channels (red, green, blue) of standard video. The highpass filter is designed to remove quiescent or slowly varying intensity values while the moving-average filter is designed to diminish noise effects. The output from each moving-average filter is weighted and squared. The weighting values were chosen by the National Television Standards Committee (NTSC) weighting to approximate the response of the human eye. The weighted and squared outputs from each channel are then summed. The square root is taken of the resulting sum. If this value exceeds a set threshold, a color change is said to have occurred. The number of terms in the moving-average were determined to maximize the signal-to-noise ratio at the output of the system. The minimum detectable color change amplitude was determined as well as the probability of false detections for various conditions.

The system was tested numerically through the use of a computer. The system was implemented in FORTRAN code. To test the system, various step amplitudes and noise conditions were input into the system. The probability of correct detection was compiled and plotted. The results show that the system can detect color changes invarious noise conditions independent of overall intensity changes. The system was also tested experimentally. A color change in Bunsen burner flame was recorded by a video camera. The resulting video data was processed using a PC-based image processing system. The results show that the system is capable of detecting color changes for color changes of different overall intensity levels.