This article offers answers to frequently asked questions (FAQs) about CalAmp Vision.

General

How does CalAmp Vision perform under different environmental conditions?

• All features work best when the road is free of mud, dirt, snow, and such and the lanes are clearly marked.
   
• If there is rain, Vision will function as expected as long as the camera can see the road ahead clearly. The camera must be in the windshield wiper–swept area (compliant to regulations as applicable in your area).
   
• If the rain is excessive, to the extent that human visibility is also compromised, certain features will not work. Because the camera works with visible light, it cannot see any better than a human eye.
   
• If there is fog and visibility is low, the camera functionality will be affected.
   
• At night, sign detection will not work unless there is ambient light, such as street lights or headlights.
   
• If there is any debris or dirt on the camera lenses, functionality will be compromised across all modules. In that case, wipe the camera lenses clean with a soft cloth. If there is any debris or dirt inside the camera, you might need to replace the camera. DO NOT attempt to open the camera and clean it.
   
• If there is condensation in the camera (as seen from its images/videos), functionality of all modules will be compromised.
   
• If the sharpness of the camera is not good (objects in images appear blurry), the artificial intelligence (AI) engine functionality will be compromised — especially for traffic sign detection, as sharpness is critical for recognizing speed limits. It is highly likely that speed limits will be missed completely, but there could be a few other errors as well.

If multiple events are triggered at the same time, will all of them have an audio warning?

When events of different categories are detected simultaneously, the audio notifications to the driver will sound out independently one after the other in sequence. There is no hierarchy or priority in the announcements, as most of the audio alerts are typically for a short duration.

If events of the same category are triggered together, they will not all have an audio warning. Each category has a silent period after an event is detected; for example, if a distraction event is detected and announced, subsequent distraction events will not make an audio alert for one minute. Forward collision warning audio alerts are silenced after the first one for 10 seconds.

Is there a self-calibration step for events to be detected?

For tailgating and forward collision warning, there is a lane-calibration step needed to analyze the geometry of the road-facing camera view. For the lane calibration to be completed, the vehicle needs to be traveling at speeds greater than 30 mph for over 10 minutes on roads that have the lanes very clearly marked. In addition, the stretch of the road should not occlude the lane markers or lines with dirt or snow. If the lane markings are not clearly visible, the calibration may take longer than 10 minutes.

Until the calibration step is completed, the device will use default values.

Distracted Driving

What is the best placement of the camera for distracted-driving detection to work well?

The camera should be mounted such that both eyes of the driver are clearly visible. It is recommended that the camera be mounted toward the driver's side (relative to the rearview mirror or center of the windshield) on the windshield, ideally in a position between the driver and the center of the windshield.

Mounting the camera toward the passenger side will result in suboptimal performance, and if you mount it too far away from the driver, the distraction module will not work at all. (This is detected automatically, and the system is muted.)

Ensure that local laws and regulations on camera mounting are followed at all times.

Note: The driver's position in the cabin is determined based on the driving side in your country.

How does distracted driving work?

Distracted driving works based on head pose relative to the direction of motion of the vehicle. The head pose is compared to the direction of motion; if the difference in orientation is significant over a set period of time, CalAmp Vision determines that the driver's eyes are off the road, and hence the driver is distracted. This type of distraction is categorized as “owl-based distraction.”

If the driver's gaze is alternating between the road ahead and below (such as if they are looking at their mobile phone), a distraction event will be generated even if the head pose is aligned with the direction of motion. In driver monitoring system (DMS) parlance, this type of distraction is referred to as “lizard eye distraction.”

This combination of head pose and gaze-based distraction works well for detecting if a driver is looking here and there, as well as looking down. Vision will catch instances in which a driver is looking down at their phone, fiddling with the controls on the dashboard while driving, or looking at something outside for an extended period.

Why do you use head pose primarily over gaze?

The driver's gaze is very hard to determine if they are wearing sunglasses, so head pose is more robust. Besides, head pose and gaze are almost always highly correlated — unless someone is purposely looking out of the corner of their eye for some time. As a result, CalAmp has found that head pose works just as well for detecting distraction.

Can distracted driving be detected at night?

Yes, this works at night because the camera has near-infrared (NIR) illumination — with the NIR LEDs coming on automatically when the ambient light is very low. As a result, the detection will also work in long tunnels.

How much is the detection affected by masks?

If a driver is wearing a mask, the performance can be diminished; some instances of distraction may not be detected as such. However, care has been taken to ensure that there are no false positives just because a driver is masked.

Can distraction be detected if a driver is wearing sunglasses?

Yes, in fact, with normal sunglasses, detection of distraction based on head pose (owl distraction) works just as well as with faces without sunglasses. However, oversized sunglasses can lower the effectiveness.

Opaque sunglasses will prevent lizard-eye distraction detection, as it requires knowing whether eyes are open or closed. Vision can detect lizard-eye distraction with normal (completely transparent) eyeglasses, though.

Do caps impact distraction detection?

They can if the cap covers the face or eyes. For distraction detection to work well, the camera should have a clear view of the face. If the cap doesn't cover the face, there will no problem.

Does the camera detect smoking and drinking?

Anything that causes the driver to look away from the direction the vehicle is moving will be detected as distracted driving. Categorizing distraction in terms of why the driver is distracted is coming soon.

Can Vision determine if the driver is drowsy or fatigued?

The camera detects that the driver is drowsy based on eye closures. Long eye closures of around four seconds will trigger a drowsiness alert.

I got an alert even though I was looking straight ahead. What happened?

If you looked away for a short duration and think you were wrongly called out for distraction, it could be due to the speed you were traveling. The distraction event is based not just on head pose but also on the speed of the vehicle and whether there is a vehicle ahead. At high speeds, if there is a vehicle in front of you, the leeway for looking around is low.

Is any calibration needed for distraction detection?

Both owl and lizard-eye distraction require calibration. The system autocalibrates itself for a certain duration (usually around one to two minutes under non-occluded, recommended camera mounting), during which it ascertains the normal driving head position and eye state of the driver. The distraction detection works based on deviation from normal driver behavior.

Drowsy Driving

How does the Drowsy Driving violation work?

The Drowsy Driving violation detects eye closures. Long eye closures of around four seconds will trigger an audio alert.

Cell-Phone Usage

Does Vision detect whether a driver is using their mobile device?

Yes, the Cell-Phone Usage violation identifies instances of the driver holding a phone to their ear.

Tailgating/Forward Collision Warning

How does the tailgating/forward collision warning (FCW) work?

A tailgating event is generated when the vehicle is driven dangerously close to the vehicle in front of it. This is designed to warn you that the driver will not have enough reaction time to avoid a collision if the vehicle ahead brakes suddenly.

Note: After the initial detection of tailgating/forward collision warning, in the next six seconds, no other events will be detected or captured.

What kinds of vehicles does the tailgating/forward collision warning work for?

The warning will work for all kinds of vehicles with four or more wheels — such as cars, jeeps, and trucks. Motorbikes are currently not supported.

What does the sensitivity parameter for tailgating do?

The sensitivity parameter controls the distance at which tailgating incidents are generated. A lower value (such as a sensitivity value of 1) triggers the tailgating event at a greater distance from the vehicle ahead than a higher value. As the sensitivity value increases, the distance at which incidents are generated reduces.

Note: Distance is only one of the parameters affecting tailgating. It is also dependent on the velocity of the vehicle — incidents will be generated at lower distances at lower speeds, and vice versa.

Why am I getting a tailgating/FCW alert when there is no vehicle in front of me?

The AI detection engine is highly precise but on rare occasions will detect a nonvehicle as a vehicle. In such cases, the driver will hear a false audio alert. The false detection might have happened due to poor visibility or because the camera saw shadows or textures/patterns on the road that looked similar to a vehicle.

Why do I get alerts even when I am maintaining a safe distance?

The safe-distance computation is dependent on the speed of the vehicle — the higher the speed, the smaller the safe-distance threshold. This variability can result in subjective interpretations. Also, the distance computation from a single camera has a tolerance of +/- 5% based on mounting height, so there may be cases where the alerts are triggered at closer distances than the intended safe distance.

Does tailgating/FCW detection work for pedestrians and cyclists?

No, currently the camera does not identify pedestrians and cyclists.

Why am I not getting tailgating alerts?

This could be caused by various issues. Try these troubleshooting tips:

• Ensure that tailgating event alerts are enabled for your fleet.
   
• Make sure that the camera is mounted correctly according to the provided instructions.
   
• Check camera health parameters to verify that the road-facing camera calibration completed successfully.
   
• Make sure that you follow the initial lane calibration guidelines required for tailgating detection. For the lane calibration to be completed, the vehicle needs to be traveling at speeds greater than 30 mph for over 10 minutes on roads where lanes are clearly marked. The stretch of the road should not occlude the lane markers or lines with dirt or snow, as that can confuse the AI. Also, in some instances, it may take longer than 10 minutes.

Harsh Cornering

How is harsh cornering detected?

A Harsh Cornering violation is reported when the vehicle’s sustained g-force in the lateral direction exceeds a predefined threshold. The sustained lateral g-force is measured over a rolling window of six seconds. The device uses inertial sensor readings to estimate the sustained g-force.

What should I do if my drivers are getting false Hard Cornering violations?

Check that the cameras are mounted per the instructions, as the detection performance is dependent on proper camera mounting.

Why am I not getting any cornering alerts?

Again, verify that the camera is mounted correctly. Detection performance is dependent on proper camera mounting.

Harsh Acceleration

How is harsh acceleration detected?

A Harsh Acceleration violation is reported when the vehicle’s sustained acceleration exceeds a predefined threshold. In other words, there is a sudden increase in velocity above a certain threshold within any given consecutive six seconds. (The sustained acceleration is measured over a rolling window of six seconds.) The device uses a combination of inertial sensor readings and GPS velocity to estimate the sustained acceleration.

Harsh Braking

How is harsh braking detected?

A Harsh Braking violation is reported when the vehicle’s sustained deceleration exceeds a predefined threshold. In other words, there is a sudden drop in velocity above a certain threshold within any given consecutive six seconds. (The sustained deceleration is measured over a rolling window of six seconds.) The device uses a combination of inertial sensor readings and GPS velocity to estimate the sustained deceleration.

I got a false harsh braking alert. What do I do?

Check that the camera is mounted per the instructions, as the detection performance is dependent on proper camera mounting.

If the false harsh braking alerts keep being repeated, verify that the duty type is correctly set for your asset/device/vehicle. If your vehicle is light duty but is configured as heavy duty, many acceleration/deceleration events will be generated.

Why am I not getting any harsh braking alerts?

Again, verify that the camera is mounted correctly. Detection performance is dependent on proper camera mounting.

Posted Speed Limit Violation (Traffic Speed Violation)

How are traffic speed violations detected?

Traffic speed violation events are triggered when the vehicle velocity exceeds the detected speed sign's limit. After the limit value is detected, the camera will check for a speeding violation at the end of the next eight seconds. If the most recent speed at the end of eight seconds is greater than the posted speed limit by the speed allowance, the violation event is captured.

When the device is evaluating whether the Speed Limit violation occurred for the duration of eight seconds, no new speed signs are detected. If the vehicle comes across additional speed signs during the eight seconds, the second and subsequent speed signs are neglected for the violation check. CalAmp Vision will use the speed limit read from the first detected sign and will report a violation when the vehicle speed exceeds its posted speed limit by the preset speed allowance value.

The eight-second interval is built into the violation check as a mechanism/opportunity for the driver to slow down after having seen the speed limit sign.

There was a speed limit sign that was read incorrectly. Why does that happen?

With AI algorithms, there is a rare chance that a sign will be misread. This is exacerbated by poor ambient lighting (such as twilight or night), backlit conditions (for example, if the sun is behind the sign when viewed from the camera), poor contrast on the sign due to wear and tear, and occlusions. In such circumstances, the events may be tagged manually. Improvements to the existing algorithms are made on a regular basis to ensure that such misreadings are reduced.

Can I set a fixed speed across the fleet above which it will be reported?

The maximum speed limit is set by default to 75 mph, but this is configurable based on duty type.

Are electronic speed signs enforced?

No, electronic speed signs aren't detected or enforced.

Maximum Speed Violation

What triggers the Maximum Speed violation?

The Maximum Speed violation is generated when the vehicle's speed exceeds the limit set by your configuration (by default, 75 mph) for a duration of six continuous seconds.

SOS (Camera)

When is the SOS (Panic Button) alert sent?

The Panic Button event is triggered by the press of the camera device’s physical SOS button.

Driver Scoring

Is there a way to base the Driver Scoring feature off the violations detected by the Vision camera?

Yes, you can enable camera-based driver safety scoring in Admin >> Driver Scoring by turning on Vision categories and then assigning a weight greater than 0 to these categories: