Quick answer: CASE cameras print on-camera analytics at three levels. Level 1 is Smart Alarm (Human & Vehicle Detection) on Value SKUs such as CS-I4XS L-2.8mm: a motion alarm that fires for people and vehicles only. Level 2 is Tripwire, Perimeter, Human & Vehicle Classification on Sharp SKUs such as CS-I5KS SD-2.8mm: drawn rules filtered by target type. Level 3 is the full AI suite on Intelligent SKUs: sixteen printed functions on CS-I8MS I-2.7-13.5mm, and a longer list on CS-I8XQ II-2.8mm that adds Visible Light Smoke & Fire Detection, Fire Exit Occupancy Detection and Face Capture. Specify by the alarm the customer must act on, quote only the printed AI row of the exact SKU, and confirm behavior per model and firmware.
Why does the AI row matter more than the series badge?
The CASE catalog groups cameras into Value, Sharp and Intelligent, and the second-generation turrets carry the II suffix. The badge tells you the tier; it does not tell you what the camera will alarm on. That is the AI row on the spec page, which changes inside a series. CS-I4KS L-2.8mm (Value) prints Tripwire, Perimeter, Motion Detection with no human or vehicle filter at all, while CS-I4KS L-2.8-12mm in the same family prints Smart Alarm (Human & Vehicle Detection). CS-I8XQ I-2.8mm carries the Intelligent label and prints Tripwire, Perimeter, Human & Vehicle Classification; its successor CS-I8XQ II-2.8mm prints a far longer list. One letter, one row, a different promise.
So: write the full SKU on the camera schedule, copy the AI row as printed into the proposal, and never translate “detection” into “recognition” or “classification” into “identification”. Everything here runs on the camera; an AI NVR adds its own analytics (AI by NVR) and search tools on top, covered in AI NVR Explained; both layers are confirmed per camera model, NVR model and firmware.
Level 1: What does Smart Alarm (Human & Vehicle Detection) actually do?
Smart Alarm is the entry level, printed on the Value SKUs CS-I4KS L-2.8-12mm, CS-I4XS L-2.8mm, CS-I6XN L-2.8mm, CS-I4US L-2.8-12mm and the dual-lens CS-I85N L-180. The camera looks for a human or a vehicle in the motion it sees and alarms on those targets; the intent is to filter out alerts for headlights, a flag in the wind or rain on the lens, the usual reason motion alerts get switched off within a month of handover.
What it is not: no printed rule set beyond the detection itself, no object-type tag a recorder can search later, no counting, loitering or object-left-behind logic. A Value SKU that prints Tripwire, Perimeter, Motion Detection instead, such as CS-I4KS L-2.8mm, gives you drawn lines and zones but fires on any motion crossing them. Level 1 fits:
- Small sites that want an after-hours alert for a person or vehicle in the yard, viewed in CASE Connect; app features and remote access options should be confirmed by model, software version and project configuration.
- Simple scenes: one gate, one lane, one door, with nothing else moving.
- Entry-tier cameras recorded on an AI NVR that runs its own rules.
Level 2: What changes with Tripwire, Perimeter and Human & Vehicle Classification?
The Sharp SKUs print Tripwire, Perimeter, Human & Vehicle Classification: among them CS-I5KS SD-2.8mm, CS-I5XS SD-2.8mm, CS-I8WS SD-2.8mm and CS-I4SV S-2.8mm, with CS-I5PS S-2.8mm adding Face Capture. Two things change. You draw the rule, a tripwire across a lane or a perimeter around a laydown area, with direction and target type, so the alarm fires only when a classified human or vehicle crosses the way you chose. And the classification travels with the event, which an AI recorder uses to let the customer search by target type; VCA Search explains that side, confirmed per NVR model and firmware.
Take CS-I5KS SD-2.8mm as the reference unit: 5MP at 3072×1728@30fps, IR to 98 ft (30 m), white light to 33 ft (10 m), 120dB WDR, IP67 and IK10, Metal+Plastic housing, −31 °F to 149 °F (−35 °C to 65 °C), 512 GB memory slot, S+265 compression. No AI row works on light the camera does not have, so keep every rule inside the printed IR or white-light distance.
Face Capture on CS-I5PS S-2.8mm is printed, but it is a project-review trigger, not a quote line: confirm use case, model, software, region and legal requirements first, as set out in Facial Recognition: When Not to Promise It.
Level 3: What is in the full AI suite on Intelligent cameras?
CS-I8MS I-2.7-13.5mm, the 8MP varifocal AI dome, prints sixteen functions: Tripwire, Double Tripwire, Perimeter, Object Abandon, Object Lost, Loitering, Running, Parking, Crowd Detection, People Counting, On Duty Detection, Video Abnormal, Audio Abnormal, Heatmap, Safety Helmet Detection and Human & Vehicle Classification; the catalog marks this set with the Advanced AI icon. The same unit prints IR to 164 ft (50 m), audio I/O 1/1 and alarm I/O 2/1, so a rule can drive a relay (a strobe, a gate hold, a panel zone) without going through the recorder. CS-I8SS I-2.7-13.5mm is the turret with the same printed list.
| Printed function | Question it answers for the customer | Typical position |
|---|---|---|
| Double Tripwire | Did someone cross both lines in sequence, in that direction? | Fence line, dock apron |
| Loitering | Who stayed in this zone longer than the set time? | ATM vestibule, rear door, fuel island |
| Object Abandon / Object Lost | What was left behind, or what went missing? | Lobby, display, tool crib |
| People Counting / Heatmap | How many came through, and where do they dwell? | Store entrance, sales floor |
| Safety Helmet Detection | Did someone enter the zone without a hard hat? | Site gate, plant floor entry |
What the second-generation II turrets add
CS-I8XQ II-2.8mm and CS-I5XQ II-2.8mm print a longer list on top of the suite: Visible Light Smoke & Fire Detection, Fire Exit Occupancy Detection, Face Capture (Up to 32 Faces / 16 Faces on the Face Detection / Capture row) and Human / Vehicle / Non-motor Vehicle / Pet Classification (Dog, Cat). The 8MP unit prints IR to 164 ft (50 m), white light to 98 ft (30 m), Red & Blue Light, a built-in mic and speaker and a 1 TB memory slot; the EW deterrence side is covered in Active Deterrence Cameras. Two guards apply. Visible-light smoke and fire detection is printed on these II turrets only (thermal SKUs such as CS-P58 T-44X print a separate thermal Fire / Smoke Detection), is a video analytic, and does not replace a listed fire alarm system (NFPA 72 in the US; CAN/ULC-S524 in Canada, verify the local code). Face Capture is confirmed for use case, firmware, region and legal requirements before it goes into writing.
Which level does the job need?
Choose by the alarm the customer must act on and the question they will ask after an incident, not by the longest list. A row they rely on that fires on shadows is a canceled monitoring contract.
| Job | Alarm the customer needs | Level | Example SKU (printed row) |
|---|---|---|---|
| Small yard, after-hours alert | Human or vehicle present | Level 1 | CS-I4XS L-2.8mm: Smart Alarm (Human & Vehicle Detection) |
| Fence line or lane with direction | Human crossing inbound only | Level 2 | CS-I5KS SD-2.8mm: Tripwire, Perimeter, Human & Vehicle Classification |
| Store door with traffic reporting | Count in and out, dwell | Level 3 | CS-I8MS I-2.7-13.5mm: People Counting, Heatmap |
| Blocked fire exit, back of house | Object or person in the exit zone | Level 3, II | CS-I8XQ II-2.8mm: Fire Exit Occupancy Detection |
| Rear lot, intervention wanted | Light, voice and alarm on a classified target | Level 3, II with EW | CS-I8XQ II-2.8mm: white light 98 ft (30 m), Red & Blue Light, speaker |
Where the customer wants rules on cameras that print only Level 1, specify an AI NVR and let the recorder run them (AI by NVR, as printed per NVR model). Where events must reach a third-party platform: ONVIF profiles supported are S, T, and G, with M on pinhole models; third-party behavior is confirmed per platform and project.
How do you commission analytics so they still work in month six?
- Bench first. Add the camera to the recorder in the shop, confirm the firmware exposes the printed row and note the version in the job file; Bench Test Before Site Day has the sequence.
- Mount for the rule. Classification needs the target large enough in the frame and at an angle that shows its shape, not a steep view onto heads.
- Draw rules where the target is lit, then set target type, size and schedule. Keep rules inside the printed IR or white-light distance; human only on a pedestrian door, vehicle only on a drive aisle; arm after hours where daytime traffic is legitimate.
- Walk-test at night and record. One technician crosses each rule in both directions while another watches the alarm list; log the result, firmware and rule screenshots in the commissioning report.
- Hand over the alert path. Show the customer the notification in CASE Connect and who receives it; features are confirmed by model, software version and project configuration.
What to tell your customer
“The cameras we specified run analytics on the camera itself, exactly as printed on each model’s specification sheet: this model alarms on people and vehicles, this one lets us draw lines and zones that fire only for the target type we choose, and this one adds counting, loitering and helmet rules. We confirmed each rule on the bench and on site at night. Face capture, smoke and fire detection and app alerts are confirmed per model, firmware and project configuration before we put them in writing, and a video analytic never replaces a listed fire alarm system.”
FAQ
Is Human & Vehicle Detection the same as Human & Vehicle Classification?
Not as printed. Value SKUs print Smart Alarm (Human & Vehicle Detection): the alarm fires when a person or vehicle is detected in the motion. Sharp and Intelligent SKUs print Human & Vehicle Classification with Tripwire and Perimeter: you draw the rule, filter it by target type, and the type tag stays with the event for later search. Confirm exact behavior per model and firmware.
Can an AI NVR add tripwire and perimeter to a Level 1 camera?
CASE AI NVRs print their own AI by NVR row, which on CS-N220 P16H-AI includes Tripwire, Double Tripwire, Perimeter and Smart Motion Detection, applied by the recorder to the incoming streams. Which cameras and how many channels can use it at the same time is confirmed per NVR model and firmware, so test it on the bench before you promise it for a camera position.
Does the full AI suite work at night?
The analytics run on the image the sensor produces, so they need a lit target. Keep rules inside the printed IR or white-light distance of the SKU, for example IR 164 ft (50 m) on CS-I8MS I-2.7-13.5mm, and test every rule after dark during commissioning. Low-light and full-color behavior is stated per SKU on the specification sheet and should be confirmed for the site.
Can I promise smoke and fire detection from a camera?
Only where it is printed: Visible Light Smoke & Fire Detection on the II turrets CS-I8XQ II-2.8mm and CS-I5XQ II-2.8mm, and thermal Fire / Smoke Detection on thermal SKUs such as CS-P58 T-44X, and only as an analytic that gives an early alert. It does not replace a listed fire alarm system under NFPA 72 in the US or CAN/ULC-S524 in Canada. Confirm use case, firmware and region before it goes into a proposal.
