I'm assuming this is for a linux or windows target system with a webcam (or mjpeg net cam) attached. I'll use opencv to monitor the camera, and tesseract will be used when OCR (character recognition) is needed. And/or OCR could be continually used to look for any text. (More CPU..) I've done several C++ OCR projects before using tesseract (and additional opencv learning layers, depending on the application requirements).