IT’S been a long wait for real world IVA solutions that are really able to take the place of human operators and I think it’s fair to say that our wait is not entirely over. Nevertheless, the latest intelligent video analytics solutions offer 24-hour monitoring and response and they are becoming increasingly discerning. 
Trouble is, a poor or inappropriate installation or application can see IVA solutions never meeting their full potential. The key is for installers, integrators, end users and consultants to understand the strengths and weakness of IVA technology and to work within its boundaries. 
There are a number of problems IVA applications will encounter and in my opinion these issues correlate nicely with those disastrous First World War battles – they are the result of the failure to establish fixed, limited objectives. So from the get-go you will need to comprehend the chosen technologies capabilities and work within this framework. 
Essentially, the IVA solution will need to avoid generating endless false alarms caused by the movement of passers-by and it will need to be capable of supplying a usable and useful image in the event of intrusion. Also important will be the degree of thought put into horizontal fields of view and depths of field. 
There will be considerations of lighting at the point of expected intrusion, the covering or elimination (or acceptance) of blind spots under and around the camera. There will be issues with the nature of the hardware – the housing and its ability to handle the environment – IP66/67 for full outdoor and IP55 under rooflines. 
As CCTV people well know, getting a good quality image without distortion or environmental intrusion of any kind over a 24-hour period is exceedingly difficult. Whether it’s loss of light, bright lights, vibration from wind, the movement of branches and trees, image attenuation caused by rain or fog, or just dusty housings, eliminating environmental noise is a tough ask. Where it gets thorny is when very definite parameters have been programmed into an IVA solution, particularly in marginal applications. 
It’s claimed that onboard analytics offer superior performance to the traditional server-based solutions because raw data is being processed rather than a compressed stream. Compression reduces the very fine detail in a scene and many cameras and compression codecs will reduce the resolution at which they record part of a scene to reduce bandwidth. The loss of detail can be significant.
Cameras with IVA processors directly in the image stream are able to undertake multiple corrections that allow them to discern movement of branches, movement of their mounting points, lighting changes and reflections, rain, dust and wildlife. 
Another worthwhile capability with IVA cameras is geo-registration. This capability allows a camera to calculate the location of the intrusion and pass the information to security teams using a Geographic Information System (GIS). You deploy GIS on big sites and when there’s an on-site team of security officers – say at a mine or a large port facility – GIS is a superb tool. 
A quality automated surveillance system gives security officers situational awareness by geo-registration of elements in a scene and by tracking targets while using this information to set security rules based on target size and velocity for more accurate outdoor video analytics.
GIS information can allow CCTV systems to automatically position a pan-tilt-zoom camera to provide an up-close view of a target. 
A good system, something like SiteLogix SightMonitor, for instance, can show the precise location and status of all cameras and targets, providing situational awareness that is both intuitive and constant. This at-a-glance multi-sensor, multi-target view displays the security posture status and increases the ability to respond to multi-target, coordinated intrusion attempts.
Cameras which employ a high degree of image processing and extract 
100 per cent of the scene information are often able to maintain a high detection rate even under low or varying light conditions. However, headlights that sweep across a visible camera’s field of view represent a special challenge, even in areas where extra illumination has been added. This is because the amount of lighting that can be added along a perimeter is small compared to the intensity of the headlights themselves.  
The use of From/To rules can be effective where headlights cause nuisance alarms. Unlike more common “trip-wire” rules, which are triggered when a detected object crosses an arbitrary line in the camera’s field of view, more sophisticated From/To rules use spatial characteristics such as size, speed, bearing and persistence.   
From/To rules are invoked when an object that maintains persistent tracking and represents the size of a pedestrian moves from one region of the scene (the “From” zone) and enters the other area of the scene (the “To” zone). In this case, the camera will determine that the detected object represents a credible threat and send an alert. This is 
accomplished with the combination of a pedestrian-sized filter along with a From/To zone to greatly reduce the likelihood of headlights causing a nuisance alert with a visible camera.  
The use of thermal imagers represents another desirable approach for solving lighting challenges. These include surveillance applications over water, where reflections would cause difficulty for visible cameras, or for surveillance applications that take place in 
total darkness.