Ring roads, squares, back streets, intersections and remote outskirts — planned with the ring-block-grid-line-point method, watched by LPR checkpoints, face-capture units and high-point PTZ, and run from one B1000 platform that scales from one district to millions of devices. Built from Uniview's official Safe City solution.
The original solution names the problems most cities already live with:
Early camera generations miss the outskirts and back streets, and at night can't resolve a face or a plate — exactly when incidents happen.
Traffic, city management and commercial systems were each built alone; when a cross-district case breaks, nobody can pull one complete picture.
Thousands of cameras produce oceans of footage, but turning it into "who and which car" still takes rooms of people watching screens.
Vehicle counts rise faster than police headcount; speeding, wrong-way and red-light running go unrecorded on most roads.
An incident is phoned in, then located, then verified, then dispatched — each handoff without live video costs minutes the victim doesn't have.
Four layers per the original design: sensing front ends chosen per scene (checkpoint LPR at city entrances, face capture at gathering points, high-point PTZ over squares, solar-powered kits in the outskirts) → the city network (access, aggregation, core) → the B1000 management platform with IP SAN storage and AI/large-model analysis → a command center with an LED wall and AR high-point view.
Simplified diagram. Real deployments follow the documented delivery guidebook: planning, environment, installation, configuration, acceptance.
Each card is a module of the original solution, told from the command center.
The ring-block-grid-line-point method turns the city into a plan: a closed ring of checkpoints at every entry road, blocks cut along district borders, grids over dense areas, lines down the main roads, and fixed points on government, schools, hospitals and banks — so every camera exists for a stated reason, not a guess.
Star-level low-light cameras cover squares, markets, back streets and old-town lanes; intrusion detection, auto-tracking and e-map linkage turn "review the tape tomorrow" into "alarm while it happens". Coverage targets 100% of key public areas, with the blind spots of earlier build-outs filled first.
Checkpoint cameras with radar capture every passing vehicle up to 300 km/h — plate, type, brand, color — and record violations from speeding to wrong-way driving. Watch lists trigger instant alarms when a flagged plate passes any checkpoint, and lane-level statistics feed the traffic department's planning.
Face-capture units at stations and gathering points detect up to 40 faces per frame; the platform indexes 26 person attributes and 22 vehicle attributes, so an operator can search "red jacket, backpack, yesterday evening" or a half-remembered plate — and replay the target's trajectory across the city map.
90+ trained algorithms handle the known cases: fights, falls, crowds, illegal parking, garbage overflow, flooded roads. On top, a large-model layer searches video by plain text ("white cargo van", "man carrying a ladder"), creates new alert rules from one sentence with zero training samples, and auto-filters most false alarms before a human sees them.
One-key alarm posts with two-way video intercom feed straight to the LED wall; the e-map pops the caller's location with surrounding cameras, an AR high-point view overlays every resource as clickable tags, and dispatch reaches officers' mobile terminals — the whole chain designed to bring response down to minutes.
Key capabilities from the official solution:
These are the equipment roles the solution is built from. Exact models are chosen per site conditions, country requirements and budget — several of our product lines fit each role, so we spec the model list after receiving your requirement list.
| Item | What it does |
|---|---|
| Dual-view PTZ (key units, squares) | Panorama and 42× detail from one head — the official list's scene camera. |
| High-point laser PTZ | 42× laser night reach from rooftops and towers. |
| Positioning system (mid-air watch) | 41× positioning unit for coastlines, rails and wide districts. |
| ITS checkpoint kit | LPR cameras, 9MP checkpoint units, radar, strobe lamps, ITS NVR — per lane plan. |
| B1000 platform servers & licenses | VM/MS/MD/DB servers, GPU analysis, per-channel licenses — sized to the city. |
| IP SAN storage | 24-bay arrays with enterprise disks; retention math decides the count. |
| Access PoE switches | Street-cabinet PoE access; aggregation and core switching quoted with the network plan. |
Browse the full product catalog — cameras, NVRs & switches →
Send your city map with district boundaries and priority zones — we reply with a ring-block-grid-line-point layout draft and a phased BOQ.
Read this before you order:
With the ring-block-grid-line-point method from the original solution. First a closed ring: every road, bridge and waterway entering the city gets a checkpoint. Then blocks: the city is cut along district borders and natural barriers, each block's entry roads controlled. Then grids over dense areas — denser downtown, sparser in quiet townships. Then lines: continuous coverage down main roads, transit stops and key routes. Finally points: government sites, schools, hospitals, banks, fuel stations and event venues get dedicated coverage. Five height layers — high altitude, rooftops, ground, underground and waterways — complete the plan.
It scales with population, road network and risk profile rather than a fixed formula — real projects run from a few hundred channels for a small city's core to tens of thousands citywide. The architecture is built for that range: one management server handles up to 10,000 channels, and distributed deployment stacks servers as phases grow. The practical answer is phased: ring checkpoints and key public areas first, then grids and lines, then AI layers — each phase usable on its own. Send your city map and priorities for a concrete phase-one count.
That is the system's defining workflow. Every capture is indexed: 26 person attributes (clothing style and colors, hat, bag, glasses, age band and more) and 22 vehicle attributes (plate, type, 180+ brands, color, even sunvisor state). An operator searches by attributes, by a photo, or — with the large-model layer — by a plain sentence like "white cargo van with a ladder on top". Results return in seconds with camera and timestamp, and one click replays the target's trajectory across the map. Watch-listed faces and plates alarm in real time at any checkpoint they pass.
The solution includes a dedicated off-grid kit: solar panels with battery storage sized for local weather, and connectivity by 4G or wireless bridging up to 12 km back to the nearest fiber point. The equipment runs from -30 °C to 55 °C, battery levels are monitored remotely, and the design cuts roughly 60% of the cost of trenching mains power to a remote site. Weak links get on-camera caching with automatic re-upload, so a dropped connection delays the evidence instead of losing it.
The reference design keeps ordinary video 30 days, key-site video 90 days and captured images 180 days — national regulations override these defaults. Recording rides on IP SAN block storage with second-level retrieval and direct camera-to-array writing. Disk protection goes beyond ordinary RAID: distributed hot-spare space rebuilds 1 TB in about 15 minutes, a failed array stays readable and new video is diverted to a backup array automatically, and integrity checks guard recordings against tampering — with role-limited, logged access on top.
Population, district boundaries and your top three problem zones are enough for a first layout and phase-one BOQ.
WhatsApp an engineer →