Member Discussion

Need IP Camera Recommendation For Sports Video Application

Need IP Camera Recommendation

I am working on a sports video analysis application which will use IP cameras to record up to four concurrent angles of a sporting event to a Windows 8.1 desktop or laptop. The purpose is essentially to create a 360 degree view of a player and use this video footage and the custom analysis tool application for evaluation and training with a coach.

The application is targeted at skill events like golf, tennis, volleyball or baseball which will primarily be held outdoors during the day or in lighted gyms. High resolution and high frame rate are the highest priority for the type of instruction we want to do; however the target market is high schools and clubs, not professional sports and pricing and reliability will also be important.

From what I have read on this site, the Samsung SNB-6004 seems like the best compromise solution; it has HD resolution, lower data rate (than the alternatives in the IPVM camera reviews, important so we can record multiple streams in real time), the 60 fps option and reasonable pricing.

Has anyone implemented a similar application or can make a recommendation on cameras that seem like a better fit? We are all software developers first and foremost and not video experts. I would really appreciate your suggestions before we start buying a bunch of cameras for testing.

Is there a reason the camera has to be IP? Normally high frame rate and high resolution mean high compression. Is your sports video app essentially the VMS or do you plan on using a COTS VMS also? H.264 or possibly MJPEG? If you expect to do 4ch frame by frame slo-mo synchronized viewing from a source of 60 fps, I would be concerned about the frame jitter making the end result unacceptable. At the very least test the camera for not only fps adequacy but also that the interval between frames is exactly what it should be.

These type of issues are dealt with in the broadcast world by very precise mechanisms (SMPTE timestripe, wordclock, blackburst) that insure all devices are transmitting phase accurate. Maybe your app isn't that demanding, but it sounds like it might be.

The application streams the H.264 video from the IP cameras to a hard drive on the laptop or computer where it is saved as one or more files, one file for each camera stream/source.

The application will display a low res preview image while capturing the video stream, but the sports analysis only happens after the video file is captured and the recording is stopped. At that time, the playback is done using a couple different options, side by side windows, overlay with transparency, or by camera angle. We haven't seen any frame jitter using the video we captured so far, but only have tested from a single IP camera.

It's not really a complicated app, but we are supporting a couple of different camera options. We have completed the implementation of streaming video from Nokia Windows WP8 smart phones and can just manually copy video from a DSLR, but we belive the most reliable and convenient option for the user will be an IP camera.

We need to find one single all purpose camera that we can throughly test and recommend to an end user, without having to support multiple vendors (if possible).

Let's assume there is no jitter/judder going on, per se. IMHO, there is a possible multi-camera frame sync concern. In my mind I'm imagining analyzing a golfer's swing in slo-mo on a long drive, from three POV's. If this isn't the type of thing you are referring to, apologies. Even at 60 fps that's still almost 20ms the frames could be out of sync between cameras. As in one stream would show the ball just prior to impact and one just after impact.

But I'm not sure that's the kinda of thing you are talking about... When you say record a baseball game to give a 360 view of one player, what do you mean exactly? How many cameras are typical? How do you keep the player in view when he runs? How close are the FOVS?

I'm not familiar with Samsung's SDK, but Axis Communications' SDK is generally regarded as the best in the business. Axis recently came out with their ARPTEC-5 architecture, which supports 60FPS at 1080P. Hopefully, you'll soon be able to buy a 60FPS 1080P camera with Axis' SDK.

I know there are Samsung distributors on this forum. Hopefully, someone will soon pop in and explain if and how the Samsung SDK compares to Axis', because the SNB-6004 is an incredible camera at an incredible price.

I am doing this already. I'm using 6 3MP Sony cameras recording Judo training at 20fps. Everything is synced together using exacqVision. If I reduce it to 1080p, then I can get 30fps, but I lose coverage area and a "hot spot" on the mat.

At 20fps, I really don't have any issues looking at the video frame by frame and can see the various aspects of what happened with no problem. More frames would be useful, but not to sure it's necessary for me. Obviously 60fps would be better, but the files are already pretty big. A 22 second clip is coming in at 22.8MB. This is just 1 camera view.

I have also used Arecont 1.3MP, 3MP, and 5MP cameras in the past. I have a 60fps Sony that I have yet to install because I was testing it for another application, but will eventually install it in this application. Everything is recording in h.264. Because of the lighting in the facility, I've had to play with the shutter speeds. That's really been my biggest challenge. If I could change out the lights, then I would have no problem getting clear stop motion when looking frame by frame, but I can't because it's not my facility.

Hope this helps.

Your example seems very similar to what we are trying to do, using higher resolution cameras rather than higher frame rate. That would be a configuration we should consider.

We do record large video files, but use tags to highlight and extract only the key scenes to share and archive and dump the rest.

Two questions: did you choose the cameras based in video quality and does higher FPS avoid the shutter speed problem?

I decided to use the Sony cameras because of their image quality, I/O input, reliability, auto-focus, remote zoom & focus, shutter control, etc. At the time, Arecont didn't have the auto-focus stuff out. I really wanted to use the 5MP because of the resolution and ability to "drill" into the image/video for detailed movement analysis, but the frame rates were too low.

When I had them at 1080p, 30fps, I lost coverage area and didn't really gain any faster shutter speed. I'll be back at the school tomorrow evening and I can try to hook up the 60fps camera and see if I clear up the shutter speeds to be faster for stop motion.

At our previous location we had flourescent lights and I didn't have any issues with shutter speeds, but at this location they're the sodium bulbs.