5.1 Video Image Detection & Microwave Radar Detection

5.1 Video Image Detection & Microwave Radar Detection

Modern traffic signal systems increasingly rely on non-intrusive detection technologies—sensors mounted above or adjacent to the roadway that do not require invasive saw-cutting into the pavement structure. For the IMSA Level III Senior Field Technician, mastering the electromagnetic theory, sensor optics, signal processing algorithms, and geometric layout constraints of these systems is critical for ensuring reliable actuated signal operation, dilemma-zone protection, and automated traffic performance measure (ATSPM) data collection.

---

1. Video Image Detection Systems (VIDS)

Sensor Optics & Image Processing Architecture

Video Image Detection Systems (VIDS) employ solid-state optical cameras utilizing Charge-Coupled Device (CCD) or Complementary Metal-Oxide-Semiconductor (CMOS) image sensors. The camera captures consecutive analog or digital video frames (typically at 30 frames per second, NTSC/PAL or IP H.264/H.265 streams) and delivers them to a local cabinet-mounted processor or edge-computing camera processor.

The optical geometry of a VIDS camera is governed by standard lens optics: $f = \frac{h_{\text{sensor}} \cdot D}{W_{\text{scene}}}$ Where:

• $f$ = required focal length of the camera lens (mm).
• $h_{\text{sensor}}$ = sensor active dimension (horizontal or vertical width of the CCD/CMOS chip, mm; e.g., 1/3-inch sensor format has a horizontal width of approximately 4.8 mm).
• $D$ = distance from camera lens to the target detection zone (ft or m).
• $W_{\text{scene}}$ = roadway field width to be captured at distance $D$ (ft or m).

The Horizontal Field of View ($HFOV$) and Vertical Field of View ($VFOV$) must be selected to encompass all target approach lanes across the entire required detection depth (from the stop line to upstream advance zones) without introducing excessive wide-angle optical barrel distortion.

Virtual Detection Zones & Feature Extraction

Within the processor software, technicians configure software-defined Virtual Detection Zones overlaid onto the video stream:

1. Stop-Bar Presence Zones: Emulate physical inductive loops (typically 6 ft wide by 20 to 50 ft long). The processor monitors pixel luminance and spatial gradient vectors within the zone boundaries. When vehicle surfaces (hood, windshield, roof) enter the bounding box, the algorithmic contrast differential triggers a phase call output to the controller.
2. Advance Count and Speed Trapping: Paired virtual zones separated by a calibrated distance ($d$) calculate approach speeds based on entry timestamps ($v = d / \Delta t$).
3. Directional & Turning Movement Filters: Vector tracking algorithms monitor the directional trajectory of edge contours, rejecting wrong-way vehicles or conflicting cross-turning movements.

The Geometric Occlusion Phenomenon

Occlusion occurs when a tall vehicle (such as a transit bus, commercial semi-trailer, or refuse truck) physically obstructs the camera's optical line-of-sight to an adjacent lane or downstream detection zone, causing two primary failure modes:

• False Actuations (Occlusion Splash): The tall vehicle's upper body projects laterally into an adjacent lane's virtual zone, placing an unwarranted false call for an unoccupied movement.
• Missed Detections (Shadowing): A smaller vehicle traveling in the adjacent lane is hidden behind the tall vehicle, preventing pixel state changes and dropping the actuation call.

Camera (Height = H)
   o
   |\
   | \ 
   |  \  Optical Line of Sight
   |   \ 
   |    \       +----------------+
   |     \      | Tall Truck     |       +-----------+
   |      \     | (Height = h_v) |       | Car       | (Occluded!)
===+=======\====|================|=======|===========|====== Pavement
   |<--S-->|<---------- D ---------->|<-- Masked -->|

The masked occlusion length ($D_{\text{occlusion}}$) cast along the roadway surface is expressed mathematically by similar triangles: $D_{\text{occlusion}} = h_v \cdot \left( \frac{D}{H - h_v} \right)$ Where:

• $H$ = camera mounting height above the roadway surface.
• $h_v$ = height of the tall obstructing vehicle (standard commercial truck $h_v \approx 13.5\text{ ft}$ / $4.1\text{ m}$).
• $D$ = longitudinal distance from the pole base to the obstructing vehicle.

Engineering Occlusion Mitigation Rules

1. Maximize Mounting Height: Cameras must be mounted at a minimum height of 25 to 35 feet (7.6 to 10.7 m) above the pavement. Mounting on luminaire extensions or dedicated mast poles significantly steepens the line-of-sight angle, drastically reducing the projected occlusion footprint.
2. Minimize Lateral Setback ($S$): Positioning the camera on a mast arm directly over the lane centerline minimizes lateral perspective distortion across adjacent lanes compared to roadside pole mounting.
3. Longitudinal Alignment: The camera optical axis must point parallel to the incoming traffic stream to avoid oblique angle distortion.

Environmental Degradations & Maintenance Protocols

Visible-spectrum VIDS performance degrades significantly under non-ideal operating conditions:

• Nighttime Headlight Glare & Blooming: High-intensity discharge (HID) and LED headlights saturate adjacent sensor pixels, projecting bright specular reflections onto wet pavement. The processor may mistake pavement glare 50 feet ahead of a car as an actual vehicle, triggering premature calls or phase extension.
• Sun Glare & Low Sun Angles: During sunrise or sunset, direct solar rays entering the lens barrel at shallow angles (within $15^\circ$ of the optical axis) wash out CMOS pixel contrast completely, causing permanent calls or total detection failure.
• Atmospheric Attenuation: Dense fog, blowing snow, and heavy rainfall scatter visible photons, dropping the signal-to-noise ratio (SNR) between vehicles and the road surface below algorithmic thresholds.
• Wind-Induced Sway: Strong wind gusts cause mast arms and poles to twist and deflect. If deflection exceeds $2^\circ$, virtual zones shift across pavement paint markings, causing false calls. Systems must incorporate Electronic Image Stabilization (EIS) and robust physical bracketry.
• Field Maintenance Requirements: Technicians must inspect camera faceplates semi-annually. Maintenance includes cleaning road grime, exhaust residue, and spider webs from optical windows using non-abrasive optical cleaners; verifying internal enclosure heater and wiper operation; and replacing internal silica gel desiccant packs to prevent internal condensation and lens fogging.

---

2. Microwave Radar Detection: Doppler vs. FMCW

Microwave radar sensors transmit electromagnetic radiation in the gigahertz spectrum, penetrating fog, heavy precipitation, direct sunlight, and darkness without performance degradation.

Microwave Doppler Radar (Velocity Detection)

Doppler radar operates by transmitting a continuous, unmodulated microwave carrier wave, typically in the X-band (10.525 GHz) or K-band (24.125 GHz). When the radiated beam strikes a moving metallic or dielectric target, the reflected frequency shifts proportional to target velocity according to the Doppler equation: $\Delta f = f_r - f_0 = \frac{2 \cdot v \cdot f_0 \cdot \cos\theta}{c}$ Where:

• $\Delta f$ = Doppler frequency shift (Hz).
• $f_0$ = transmitted carrier frequency (Hz).
• $v$ = vehicle velocity relative to the road surface (m/s).
• $\theta$ = angle between the radar beam transmission axis and the vehicle velocity vector.
• $c$ = speed of light ($3.0 \times 10^8\text{ m/s}$).

The Stationary Vehicle Blind Spot

The critical engineering limitation of pure Doppler radar is that stationary vehicles produce zero Doppler shift ($\Delta f = 0$). When a vehicle comes to a complete halt at a red light, a continuous-wave Doppler sensor ceases to detect it. While firmware can hold an artificial call via software "extension timers" (typically holding a call for 10 to 60 seconds after motion ceases), true static presence detection is fundamentally impossible with Doppler radar alone. Consequently, Doppler radar is deployed primarily for advance dilemma-zone speed monitoring or dilemma-zone extension, never for primary stop-bar queue presence.

Frequency Modulated Continuous Wave (FMCW) Radar

To achieve true static presence detection alongside range, speed, and angle tracking, modern intelligent transportation systems deploy Frequency Modulated Continuous Wave (FMCW) radar. FMCW sensors operate primarily in the K-band (24.0 to 24.25 GHz) or W-band (76 to 81 GHz).

Frequency
   ^
   |      Transmitted Chirp (Sweep Bandwidth B)
   |     /|          /|
   |    / |         / |
   |   /  |        /  |
   |  /   |       /   |     Received Chirp (Delayed by t_d = 2R/c)
   | /    |  fb  /    |
   |/     | <-->/     |
   +------+----+------+---> Time
   |< T_s >|

FMCW Operating Principles

1. Linear Frequency Chirp: The radar synthesizes a carrier wave whose frequency is modulated linearly over time across a sweep bandwidth ($B$) during sweep time ($T_s$): $\frac{df}{dt} = \frac{B}{T_s}$
2. Range Calculation via Beat Frequency: The signal reflects from a target at range $R$ and returns after round-trip propagation time $t_d = 2R / c$. The received signal is mixed with the currently transmitting signal, producing an intermediate Beat Frequency ($f_b$): $f_b = \left( \frac{B}{T_s} \right) \cdot t_d = \left( \frac{B}{T_s} \right) \cdot \left( \frac{2R}{c} \right)$ Solving for target range ($R$): $R = \frac{c \cdot T_s \cdot f_b}{2 B}$ Because $f_b$ depends strictly on round-trip distance, an FMCW radar accurately measures distance to completely stationary objects. A stopped vehicle at the stop bar produces a constant, non-zero beat frequency, ensuring permanent, uncompromised presence detection.
3. Radial Velocity Resolution: Velocity is measured simultaneously by tracking the phase shift of consecutive chirps across a coherent processing interval (Doppler FFT).
4. Azimuth Angle Determination: Multi-channel Digital Beamforming (DBF) or Multiple-Input Multiple-Output (MIMO) antenna arrays measure the phase arrival difference ($\Delta \phi$) across physical antenna elements separated by distance $d_{\text{ant}}$: $\theta = \arcsin\left( \frac{\lambda \cdot \Delta \phi}{2 \pi \cdot d_{\text{ant}}} \right)$

Advanced Intersection Tracking Capabilities

High-resolution 77–81 GHz FMCW radar provides multi-lane, multi-object tracking:

• Simultaneously tracks up to 64 to 128 individual vehicles across 8 to 10 approach lanes.
• Distinguishes vehicles longitudinally with range resolution down to 0.5 feet (0.15 m) ($R_{\text{res}} = c / 2B$).
• Tracks vehicle trajectories from 900 feet (275 m) upstream down to the stop bar, dynamically calculating Time-to-Stop-Bar ($TTSB$) to deliver dynamic dilemma-zone protection.