The Business Case for Computer Vision in Transportation

Fleet management platforms enable companies to monitor and manage fleets of five or more vehicles while improving safety and reducing operating costs. Now, computer vision (CV) technology is being applied as part of advanced driver assistance systems (ADAS) to provide real-time alerting and analysis of data captured by dash cameras. It also creates actionable events as shown in the illustration below.

Figure 1: A Functional Overview of Computer Vision in Fleet Management

Fleet management uses a variety of components to implement safety and monitoring via CV:

Dash cameras: Continuously captures video of the inside and outside of the vehicle, however only video related to interesting events is stored to conserve space. Some events, like an impact, trigger CV techniques to extract information from the videos.

Advanced driver assistance systems: Helps drivers avoid accidents by providing audible and visual alerts to real-time events and uses image processing extensively.

Live streaming: Enables real-time monitoring by continuously streaming to the cloud, where techniques like sentiment analysis are used to annotate footage of interest and determine positive, negative, or neutral sentiment for each moment.

Actions: Responds to events like sudden braking by applying CV-powered video analysis to take appropriate action, for example making sure the driver is unharmed. Personnel monitoring the streaming video footage can telephone drivers to check on them. Video can also be used for more mundane maintenance, such as to determine if a vehicle needs cleaning, or if there is a cracked windshield.

Enabling CV-driven solutions are connected dash cameras that provide a 360-degree field of view both inside and outside the vehicle and can stream over radio-area networks (RAN) such as 4G LTE or 5G. Also required: vehicle access via API or specialized hardware plugged into an on-board diagnostics (OBD) port and able to communicate via a RAN.

CV in transportation protects drivers, vehicles, and pedestrians. ADAS uses sensors, radar, light detection, and ranging (LiDAR) for depth perception. It also uses cameras and embedded image processing software to improve driver and pedestrian safety.

Dash cameras capture high-definition footage of incidents and notify drivers via in-cab alerts related to distracted driving situations. In addition, incident footage can be used for training drivers.

The benefits of computer vision in transportation include:

• Increase safety, reducing insurance claims by 15% to 30% due to fewer accidents, better driver education, and improved driving habits.
• Reduce fuel costs by up to 20% through improved vehicle monitoring via vehicle telematics and alerts, and adoption of good driving practices like maintaining optimal speeds.
• Improve driver behavior in events like harsh braking, distraction, and speeding by up to 75% due to training and monitoring.

Fleets of vehicles are used in a large number of industries, which can make use of CV to expand the capabilities of their fleet management systems. These industries include logistics, taxicabs, freight, last-mile delivery, rental cars, non-emergency medical transport, and university and company vehicles.

Transportation workflows are complex and fleet management is no different. CV increases the safety factor for these workflows and is used in the following scenarios with fleet management:

CV is used in the following scenarios with fleet management:

• Driver-assisted security: ADAS alerts the driver or automatically takes corrective action in real time. Functions include pedestrian detection/avoidance, lane departure warning/correction, traffic-sign recognition, automatic emergency braking, and blind-spot detection.
• Warn drivers based on vehicle monitoring: The system notifies the driver when distracted driving, speeding, or tailgating (based on dash-camera footage and vehicle telematics) is detected. Central monitoring station personnel notifies the driver via back channels.
• Record footage: Dash cameras automatically capture and transmit footage related to an incident, like sudden braking or impact. The video can be used to support corrective action or filing of an insurance claim.
• Training and coaching: Video footage can help train drivers, leveraging sentiment analysis and emotion detection to assess driving habits.
• Theft prevention: Live stream footage and geo-tracking enable central monitoring of the vehicle, which can be remotely disabled in the case of theft.
• Improved maintenance: Dash cams can detect a dirty vehicle, which can be flagged for cleaning. They also detect safety issues like broken headlights or a cracked windshield, initiating a notification that eliminates the need for manual inspections and improves maintenance workflows.

The alternative to CV in fleet management is supporting vehicle operations manually or in a semi-automated way. Semi-automation implies fleet management software operating without vehicle telematics and CV. This usually results in increased costs for vehicle maintenance, an inability to guarantee HOS compliance, and lack of driver and pedestrian safety controls that result in higher accident rate.

Organizations can expect to spend from $10 to $50 each month per vehicle for the camera, CV technology, and video connectivity to the cloud.

Fleet owners incur a one-time cost to purchase a front and rear dash camera or a 360-degree dash camera that ranges from $200 to $500 for each vehicle, plus implementation fees. They also incur a one-time system installation fee of $1,200 to $2,500 to retrofit ADAS into the vehicles.

Other recurring costs include baseline monthly charges for a fleet management system that manages vehicles, telematics, and drivers, as well as additional storage costs for video per GB used, charges for extra retention periods of video storage, and ongoing training costs per driver.

The largest risk factors are security, system performance and operations, and personnel training. Security risks are cyber intrusion of cameras, unauthorized access to telematics data, and privacy. System performance risks are incorrect identification of objects and volume of transactions. Operations risks are loss of regulatory compliance and inadequate business processes. Lack of driver training can impact fleet operations and cause higher insurance or maintenance costs. The use of video recording may also negatively impact driver recruitment since they might not want to be recorded all the time.

Any CV application should also consider the following risks:

• Ethical: Does the solution violate business ethics when using data for incorrect purposes, is the data skewed on demographics, or is the application legal?
• Economic: Does the solution have a potential for lawsuits or could there be an impact to the reputation of the organization?
• Cultural: Is there cooperation between drivers and other teams to prevent undesired outcomes that go undetected for long periods of time?

The path to adopting CV in transportation for fleet management depends on the scale and scope of the implementation. For general guidance, the following roadmap is highlighted in our 30/60/90 day plan:

30 Days: Identify and scope. Identify use cases, vendors of dash cameras, vendors of fleet solutions, and vehicle telematics to track. Determine fleet size and desired growth over three years. Classify fleet vehicles by brand and model so that deployment plans account for hardware installation of devices, ADAS solutions (if needed), and cameras.

60 Days: Trial deployment and define workflows. Line up vendor trials, conduct POCs, and understand companion features and costs. Determine if additional modules for vehicle telematics are needed and if the vendor or partner provides these modules. Categorize actions for each workflow using appropriate vehicle telematics. Share workflows with vendors. Define key performance indicators (KPIs) for the workflows. Work with a legal team to craft employee agreements for drivers and plan training programs for drivers.

90 Days: Deploy and extend. After deployment, Purchase the fleet management platform, deploy hardware, and create sample workflows for a few vehicles before expanding rollout of sample workflows to all vehicles. Draw out plans for custom development, hardware deployment, and implementation of fleet management workflows for all vehicles. Prepare Draw out plans to go live with all production workflows. Commence training for drivers. Define a plan to collect and access telematics and track metrics to determine KPIs.

• Key Criteria for Evaluating Mobile Edge Solutions v1.0