Television data collection and return technologies are one of key technologies in television secure broadcasting system, TV video content surveillance, TV program copyright protection, and client advertisement broadcasting. In china, the dominating methods of TV video content surveillance are manual tape recording and whole TV program Automatic Return. Manual method costs too much, whole TV program return method needs lots of net bandwidth and storage space. This paper proposes a new method of television data collection and return technology, video field is extracted from continuous video and coded at frequency of about one field per second, in other words, one field is extracted from continuous fifty fields of original video for PAL TV system, extracted frame can be coded by all means, for example JPEG2000, or intra mode code of H.264 or MPEG2. TV programs whose content and topic change most frequently are news and advertisement program, which may change topic in five to ten seconds, so extracted sequences hold the same topic and content and enough information with original video for TV program content surveillance application. The data quantity of extracted sequence is about 3 percent of the original video program, which will save large quantity of network bandwidth and storage space. One hardware implementation method of this technology based on embedded system is proposed, the TV Field Extractor, which circularly extracts images from target TV program, uses high-performance compression algorithm for image compression and stores the final output sequences of stationary images on the hard disk, or transmits these sequences to the monitoring center via network. This method evidently reduces device cost, network bandwidth and storage space, which can be widely adopted in TV program content surveillance and TV secure broadcasting system.