The present invention provides a SIP-based network video surveillance system and method, which relates to the application field of network video surveillance. The network video surveillance method comprises the steps of: receiving a request message from a client by a call control server (CCS), the request message comprising an ID of a Premises Unit (PU) and a channel ID; forwarding the request message to a center management server (CMS) by the CCS for access authentication; transmitting the request message to the PU by the CCS after the request message passes the access authentication; responding by the PU an acknowledgement message containing video codec information to the CMS via the CCS; forwarding the acknowledgement message containing the video codec information and an access path to the client by the CMS, then establishing, by the client, a video stream connection with a network element directed by the access path on the basis of the acknowledge message. The present invention uses SIP in conjunction with other protocols to build a complete video surveillance architecture, improving the extensibility of a platform and makes the whole system more stable and reliable.