City-Wide AI Engine for Multi-Camera ANPR Trajectory Tracking and Urban Traffic Analytics
Bharat Electronics Limited · Transportation & Logistics · Software
The trajectory view is the memorable part, so build that and the fuzzy plate matching that makes it robust — but be careful about the ninety percent claim, because stating a number invites being tested against it.
What it actually is
Cities have hundreds of number-plate cameras but each one works in isolation, so nobody can follow a single vehicle across the network or extract city-wide movement patterns. The ask is an engine that reads plates accurately in bad conditions, reconstructs any vehicle's full journey across cameras, and turns the aggregate into traffic analytics. The description sets a specific accuracy bar.
What to build
A three-part platform: a plate recognition engine reading Indian number plates across varying light, weather, viewing angle, motion blur and physical plate damage, reporting a per-read confidence; a trajectory reconstruction module that links reads of the same plate across geographically distributed cameras into an ordered journey with timestamps, direction and inferred route drawn on a GIS map; and an analytics layer aggregating all reads into traffic density measures, origin-destination matrices, congestion bottleneck identification and live movement heat maps for the city authority.
Smallest thing that wins the room
Search a specific number plate and watch its complete journey reconstruct across six camera locations on the city map in time order, with the gaps between cameras interpolated and each read's source frame and confidence shown alongside.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 20% of the 226 · #180 of 226 by expected field
Busier than most. Expect several teams to arrive at the same obvious solution.
Why: company-sponsored statements drew the smallest fields of all.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
3/5The trajectory reconstruction is genuinely valuable and demos beautifully, but you have committed to a stated ninety percent accuracy figure that a BEL judge will ask you to evidence on hard cases, and without a real camera network the cross-camera half is demonstrated on a deployment you invented.
Feasibility
3/5Plate detection and recognition have public Indian datasets and strong open models, but the description's stated bar of over ninety percent accuracy across motion blur, poor weather, angled shots and damaged plates is a demanding target, and you have no real multi-camera network so the trajectory half runs on simulated deployments.
Innovation scope
3/5Plate recognition is a mature commodity task and trajectory linking is straightforward once reads exist, so your creative room is mainly in handling uncertain and partial reads when reconstructing a journey.
Clarity
5/5The description names three core functionalities, specifies the accuracy threshold and the adverse conditions it must hold under, and enumerates the four expected platform components, leaving the requirement unusually precise.
Effort
MassiveA robust recognition engine, a cross-camera linking system and a full analytics layer are three substantial builds, and reaching the stated accuracy on degraded plates alone consumes serious time.
Demo-ability
EasyWatching a vehicle's journey assemble itself across a city map from scattered camera reads is immediately compelling and needs no explanation.
In its favour
- Green flag: The trajectory reconstruction across a map is a strong, distinctive visual that separates this from the many plain plate-reader submissions
- Green flag: Fuzzy matching on partial or misread plate strings is a real technical contribution that makes trajectories robust where exact matching would fail
- Green flag: The accuracy bar is stated, so you have a defined target and a clear structure for your evaluation section
- Green flag: Public Indian plate datasets plus your own captured footage give you an achievable path to a defensible test set
Against it
- Red flag: The description commits you to over ninety percent accuracy across motion blur, weather, angles and damaged plates — quote that figure and a judge will hand you the hardest cases to prove it
- Red flag: There is no real multi-camera network available, so the cross-camera linking is demonstrated on a simulated deployment you constructed yourself
- Red flag: City-wide vehicle tracking is a surveillance capability, and a judge may reasonably ask about access control, retention and misuse — have an answer rather than being surprised
- Red flag: It overlaps the ANPR and tracking portion of SIH26124 from the same organisation, so be ready to explain why a dedicated engine is the better answer
What you will be writing
- YOLOv8 plate detection + PaddleOCR / TrOCR recognition
- Indian plate datasets (IDD, ALPR benchmarks)
- Plate-string fuzzy matching for partial reads
- PostGIS trajectory reconstruction and map matching
- Kepler.gl / deck.gl OD flow and heat maps
- Kafka multi-camera stream ingestion
- Automatic number plate recognition
- Urban surveillance analytics
- Spatiotemporal tracking
Prior art to read before you start
multi-camera vehicle trajectory tracking · plate recognition under adverse conditions · origin-destination traffic analytics
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.