Successful video analytics deployment depends on two things: understanding when and why to use it and then deploying successfully. An in-depth understanding on how video surveillance will be used will directly impact how successful the deployment will be. Here are 10 questions to ask to ensure successful video analytics deployment:
1. What does the customer want from the system?
This is number 1 for a reason, understanding and managing customer expectations is the most important step to ensure that the final delivered system performs as expected. Reviewing and documenting customer expectations in terms of coverage area, detail capture and response protocols will help in the selection of proper cameras and equipment to satisfy requirements.
2. What types of rules will be used?
Simple rules, like detecting a person present in a given area offer lots of flexibility in terms of camera placement. More specific rules, like direction of travel, might restrict camera placement options. For example, movement of a person left to right in a field of view is going to be more visually apparent than a person walking directly to a camera’s field of view. Both deployment options are acceptable, but the camera positioned to see the object traveling left-to-right will tend to react faster or require less movement by the object in order for the rule to trigger.
3. What are the high-value assets and location?
The return on investment of securing high-value property is immeasurable. In many cases protection can be established more cost effectively if there is flexibility in the storage location of particular assets. Grouping high-risk items in proximity to each other can potentially reduce the requirement to cover an entire property, making the system more cost effective or affordable. This also makes monitoring easier. Sometimes the best designs are a combination of technology and recommendations to make the physical site itself easier to secure.
4. What is the lighting performance?
Many times analytics are being used for perimeter protection and the risk of loss can be higher at night. Cameras see light and more light tends to make for a better image. It is important to know how the site's lighting is set up and to make sure you're getting good light coverage where you expect advanced analytics to perform. Some cameras may offer things like "sensitivity up," or "moonlight mode," which can lead you to believe an insufficient lighting arrangement can be made workable. However, these features often result in a tradeoff between a grainy image or a slow shutter speed, resulting in a video feed that cannot be accurately analyzed. Understand the illumination requirements of your equipment so that you can ensure on-site lighting meets these requirements.
5. What are weather conditions on site?
Extreme weather such as bright sunlight, glare, wind and rain/snow can impact the accuracy and reliability of surveillance systems. If customers have surveillance needs where extreme weather is common, it might make sense to position cameras closer to the target areas (reducing the chances of weather interference in the shot), add additional lighting or stabilize mounting poles to reduce camera sway or movement in high winds.
6. What are possible obstructions?
In order for analytics to "see" objects, they need to actually be able to see the object! Sometimes simple things like trees or bushes can block part of a shot, other times it's vehicles parked at a car dealership, or piles of materials at a scrap metal recycling facility. A single camera can cover several hundred feet of perimeter, so it is important to make sure your coverage isn't cut short by an obstruction. Camera placement for analytics deployments will take the entire site into consideration and sometimes require moving a camera location to get the best view.
7. How is the system calibrated and maintained?
Systems that require extensive manual calibration and setup will be more difficult and expensive to maintain over time. Many times users only realize their system needed additional calibration after a theft or event has occurred!
8. What are the camera placement restrictions?
Restrictions on camera placement locations can affect the design. Sometimes these restrictions are governed by other site safety criteria, as in refinery or electrical substation sites, other times they are simply a cost trade-off. Will spending thousands of dollars to customize a placement or location be worthwhile, or can an extra camera be mounted elsewhere to give the same coverage? It’s important to know beforehand if there are any site-specific criteria that will limit camera placement options.
9. What’s the Internet connectivity at the protected premises?
Video analytics are typically deployed in scenarios where there is some kind of live monitoring, either by a third party central station or end-user. In order for events to reach the operators in a timely manner and for those operators to view video or interact via audio talk-down, you will need Internet connectivity at the site. Alarm sending and live video viewing rely on the UPSTREAM bandwidth (data going OUT of the site), which is usually lower than the downstream bandwidth and frequently glossed over in ISP specs. Ensure that the upstream bandwidth can sustain at least 500Kbps, and that the data usage contracts do not limit usage. A typical site will use about 4 to 8GB per month in data transfer, assuming six cameras and a moderate amount of activity and live viewing. Static IP’s for the Internet connection are not strictly required, but will make life easier. Many ISP’s offer a static IP option for around $10 or $20 per month; it’s a wise choice if that option is available.
10. What options are available to handle false alarms?
A good analytics system is able to do more than simply classify objects as “yes” or “no.” Ideally we all strive for zero false alarms, but there are a wide variety of scenarios encountered and better/more accurate configuration options can help keep false alarms low. You should have precise control over Regions of Interest (the actual part of the field of view where the analytics are applied), the ability to distinguish objects by type (eg: person vs. vehicle), and the ability to have different rules and regions active based on the time of day or the day of the week. This will help ensure the system is not inadvertently triggered by nuisances like animals or by employees during normal working hours.
