SSM-3600 high-throughput video intelligent detection system
SmartSys-Media-3600 high-throughput video intelligent detection system，to high-throughput video intelligent processing chip as the core，integrated video matching、 logo recognition, pornography、video enrichment、video transcoding and low latency video and other basic functions for many applications to provide a high-throughput、 high-density,、low-power service platform.
With the rapid development of Internet video applications，the amount of video data has increased dramatically. The use of existing server clusters to provide video detection services, faced with many challenges: the amount of video data surge caused by a single node processing capacity; demand for heterogeneous system, resulting in functional overlap and waste of resources; server room space is not enough, power Surge.
SmartSys-Media-3600 high-throughput video intelligent detection system provides a solution to the above problems.
Firstly, the video feature library is extracted from the mass video and the high-level semantic features, and then the characteristics of the sample video are extracted and matched in the feature library to detect the video in the video library with the same or partial video of the sample video.
l Detection rate is high, the correct rate of 99.99%;
l Video detection speed, single frame processing time is less than 10ms;
l Video detection sample library capacity, support tens of millions of catalog index.
l Video search positioning: through the sample video or sample pictures in the video library to find the original video or source, allowing users through the most simple and direct search, accurate search for the required video clips;
l Video screening and copyright protection: to quickly check the video frequency characteristics, and in the copyright video feature library by contrast, once finding the high similarity with the copyright video, you can determine the suspected pirated video.
Detect the content, location and changes of the television station logo in the video.
l Extraction accuracy is high, the accuracy rate is higher than 99.9%;
l Match the high accuracy, the correct rate of 99.9%;
l Extraction, detection speed greatly, single frame processing time less than 30ms;
l Large capacity of comparison Library, can support 1000 logo sample index detection.
l Applied to real-time monitoring of broadcast and television stations to prevent television programs in the broadcast process was illegal spots and interference;
l In the video site, automatically detects the source of video resources uploaded by users ,manage video resources effectively;
l Supervise foreign video to prevent illegal or unhealthy video inflows;
l Verification of video copyright infringement.
Automatically searching and detecting videos or clips that have pornographic scenes in video assets.
l The accuracy rate of detection arrived at 95%;
l Single frame processing time is less than 200ms.
l In the large video site, the daily on the million new video upload automatically when the detection and found violations in a timely manner, from the source control of the emergence of pornographic video；
l In the operator gateway, the flow of video through real-time monitoring, once finding violations, which are cut off in time to avoid large-scale dissemination of pornographic video.
Extract the moving target in the video through intelligent analysis, and supplemented by lower bit rate, reduce the image, multi-frame merging and other means to retain the key content under the premise of reducing the video storage space, shorten the playing time.
l Storage space compression ratio of up to 100 to 1;
l Play time compression ratio up to 200 to 1;
l Processing speed of up to 4Gbps.
l Smart City: In the video surveillance, urban alarm, intelligent transportation, wisdom, education, medical and other related video services, the use of video enrichment can improve the efficiency of retrieval and reduce storage costs.
Converting multimedia from different sources to multiple resolution, different bit rates, and multiple target files or data streams in different formats;
l Support for rich video formats;
l Support multi-format file output (one decoding multiple coding);
l support image zoom, multi-channel video splicing and picture-in-picture;
l Support for adding watermarks and auto screenshots;
l Internet video on demand, live game；
l Multimedia post-production and processing；
l Video conferencing。
Low latency video
The end-to-end delay from acquisition to presentation is the duration of one frame or two frames.
l Coding delay less than 10ms;
l The delay is less than 60ms from end to end, human almost feel.
l Cloud game: game application running on the cloud server, the terminal only provides network transmission, video decoding and human-computer interaction. The use of low-latency video technology to achieve the game control operation and screen display between the synchronization;
l Virtual reality VR: the use of computer simulation to produce a virtual world of three-dimensional space, provide auditory, visual and other sensory simulation, so that users as immersive in general, timely and unlimited observation of three-dimensional space things. Low latency video technology is the key to ensuring a user experience.
Broadcast and TV Station, TV film-make, Media Company, Video site, Video Resource Provider, Propaganda department, Public Safety Department