Imagine ripping along at 25 to 30 knots in the dark, in a big seaway, singlehanded aboard a 60-foot offshore racing sailboat in the nonstop around-the-world Vendée Globe race. Land and help are hundreds of miles away. Sleep is one of your most valuable currencies, but commercial vessels, fishing boats and whales also transit these waters. Trusting the big-ocean theory while you get some shut-eye can be risky business.

Optical-based collision-avoidance systems are a solution to this problem. One example is Sea.AI (née Oscar), which was developed in 2018 to help keep these kinds of sailors safe. Flash-forward seven years, and this type of technology is protecting boaters of all stripes, with numerous brands on the market and companies competing to advance the systems in various ways.

Optical-based collision-avoidance tech is an offshoot of automotive-based, advanced driver-assistance systems. This technology is quickly becoming an invaluable safety net, alongside radar and the automatic identification system, aboard well-equipped yachts. Elements of this technology are also critical for enabling assisted and autonomous docking and navigation systems. Contemporary systems alert captains of potential collision threats, with AI’s evolutionary curve suggesting more to come. Much like a car’s ADAS, this tech could soon also be standard kit aboard boats.

Most optical-based collision-avoidance systems have one or more cameras, an AI-enabled black-box processor and a display. Systems can include a daylight camera with low-light capabilities or a thermal-imaging camera, or both. The processor typically contains a library of annotated images that depict, for example, a vessel at sunset, a buoy in waves or a partially submerged container. The screen, which can be dedicated glass or a networked multifunction display, presents visual and audible alarms and real-time video imagery of any camera-captured targets.

The camera’s video is fed through the processor using AI computer vision and machine learning. It essentially lets the processor “see” through the camera. The processor then compares the camera’s real-time video feed with its imagery database, or it uses its knowledge of how to identify targets based on its annotated imagery database to identify nonwater objects in the camera’s field of view—a sailboat in the fog, for example.

“Our database contains more than 20 million objects in different scenarios, like sea states, weather conditions, geographic locations,” says Christian Rankl, Sea.AI’s chief technical officer. “It’s key to have a database with a wide range of objects and scenarios to build a highly reliable collision-avoidance system.”

Once the system has identified an object, it tracks it and calculates the real-time distance and bearing to the object, as well as a safe course (depicted on the display) around it.

The math isn’t trivial, says Sangwon Shin, vice president of recreational marine for Avikus, a subsidiary of HD Hyundai that specializes in autonomous navigation: “The hardest part about creating a collision-avoidance system is calculating the distance.” Factors include the boat’s pitch and roll, plus the marine environment’s diverse conditions. A boat’s distance from an object and its velocity also factor into calculating an avoidance path.

This all unfurls almost instantaneously with Avikus’ Neuboat Navi system. “It takes about 20 to 30 milliseconds,” Shin says about the time frame required to identify an object. The system, which uses an electro-optical camera and a lidar sensor to measure distance, recalculates this 10 times per second to ensure accuracy. “Sending the alarm to the boaters takes about 100 to 200 milliseconds,” Shin adds.

Other systems also offer processing times that are lightning-fast. Phil Bourque, Sea Machines’ vice president of global sales, says his company’s AI-ris system has latency of less than 0.25 seconds at full 4K resolution. “So, it does a lot of thinking very quickly.”

But speed is only one necessary component of these systems. They also have to minimize false alarms. Rankl says Sea.AI continuously refines its AI model by analyzing scenarios where it performed poorly. “It’s crucial for the AI to accurately distinguish real threats from benign objects.”

Sensor payload is another area where evolution is occurring, beyond hardware, software and AI models.

“While optical and thermal sensors are highly effective in detecting various floating objects, they, like all sensors, have limitations,” Rankl says, noting that these limitations could be addressed by integrating radar, AIS, lidar and sonar. “Our research department is actively evaluating the value these sensors can provide to our customers and how they can further enhance their safety at sea.”

Bourque agrees, noting that Sea Machines is working to integrate AIS and radar into AI-ris. “We certainly see the demand for the fusion of computer vision, radar and AIS,” he says.

Another important integration involves displayed cartography and data overlays. Anyone who cruises with radar and AIS is familiar with how multifunction displays can overlay AIS targets and radar data atop vector cartography. To that end, Sea.AI recently partnered with TimeZero to display targets detected by Sea.AI’s Sentry system atop TimeZero’s TZ Professional navigation software. “We are actively working toward integrating our machine vision with other platforms as well,” Rankl says.

Sea.AI isn’t alone in this thinking. Avikus’ Neuboat Navi presents camera-detected targets in its real-time head-up display, and Sea Machines’ SM300 autonomous command and control system displays camera-detected targets atop cartography.

The trick, of course, will be getting optically detected targets onto mainstream multifunction displays, but multiple sources say this is already in the works.

Accurately assessing the future of optical-based collision-avoidance systems is a tougher ask.

Bourque says the next five years should see these systems mature and progress—much like the ADAS performance curve. He also says today’s refit customers will want this technology to come factory-installed aboard their next yachts, necessitating that designers and builders allocate physical space for these systems.

In addition, Rankl says, optical-based collision-avoidance technology will become a standard feature on boats, akin to radar and AIS. He sees low-Earth-orbit satellites such as Starlink playing a big role with their fast, global connectivity.

“This will enable the development of large vision models specialized for maritime use,” he says. Rankl also predicts that the rise of AI spatial intelligence, which allows AI models to understand and interact with geographic information, will let collision-avoidance systems better predict the movements of detected targets based on their positions and trajectories.

“Over the next five to 10 years, we expect multimodal systems that integrate data from all available boat sensors—cameras, radars, AIS, etc.—into a unified AI acting as a 24/7 co-skipper,” Rankl says.

Shin agrees but is more bullish about the time frame, which he puts at three to five years. “This technology will be developed in a way that combines multiple sensors and provides more accurate information,” he says. In five to 10 years, he adds, a single piece of hardware will provide “all the necessary data for collision avoidance.” As far as autonomous docking and navigation, Shin says: “We do not aim only to give situational awareness and provide suggested collision-avoidance routing. Our ultimate goal is to provide [an] autonomous system for boats, which is only possible with accurate distance calculation.”

Sea Machines is also integrating its optical-based collision-avoidance system with autopilot and engine controls to enable autonomous decision-making. Sea.AI is exploring options and applications for its technology.

As with all technologies, optical-based collision-avoidance systems aren’t without their high and low tides. On the positive side, these stand-alone systems add significant safety margins and don’t rely on signals transmitted from other vessels. Conversely, all technologies add cost and complexity, and false alarms can trigger unnecessary stress.

While today’s optical-based collision-avoidance systems offer a sea-change advancement over trusting the big-ocean theory, it will be fascinating to see what future directions the technology takes. Either way, there’s no question that technology which began as specialized equipment for racing sailors is already having a massive impact on the wider boating world.

Evading Other Emergencies

In addition to spotting potential collision targets, optical-based detection systems can be used to locate and track a crewmember who has fallen overboard. Since these systems don’t rely on incoming AIS signals or radar returns, they can be key for detecting, identifying and tracking possible piracy threats.

Nautical Nightmare

A crewmember overboard is one of every captain’s worst fears, but the same camera systems that can help avoid collisions can be used to locate crewmembers in distress.

The post Optical-Based Collision-Avoidance Tech appeared first on Yachting.

In seventh grade, Matthew Zimmerman dreamed of playing in the rock band Metallica. He picked up an electric bass, and by high school, he was playing upright bass. In college, he joined the University of Rhode Island’s jazz band and his professor’s house band—experiences with steep learning curves. Zimmerman was on the six-year plan, “adding majors.” He graduated with bachelor’s degrees in French and German, another bachelor’s in ocean engineering, and several key relationships.

“There’s lots of connections between engineering students and music,” he says, referring to the mathematics of music. “You don’t notice a great bass player unless they mess up. Obstacle avoidance is like that: Best case, you avoid catastrophe.”

His path to becoming FarSounder’s co-founder and CEO took root during those college years, when he began working with Jim Miller, a professor in the Department of Ocean Engineering. A graduate student had been working with Miller on a three-dimensional forward-looking sonar project. When that student left, Zimmerman took over. The project earned some publicity, and they got a call from an oil company.

“I either had to find a job or turn the FLS project into my job,” Zimmerman says while sitting at the conference table in FarSounder’s Warwick, Rhode Island, headquarters. Their technology wasn’t a good fit for slow-turning oil tankers, but the call inspired confidence. “Our motivation was to help mariners avoid hitting whales and rocks.”

Zimmerman and Miller formed FarSounder in 2001 using private funding from family, friends and angel investors. Miller kept his position at the university and joined FarSounder’s board of directors. Given that Zimmerman was 24 and more interested in developing sensor technology than in business operations, the company hired—and then fired—a CEO. Enter Cheryl Zimmerman, Matthew’s mother. She had helped form the company, and her business experience made her an ideal replacement as FarSounder’s CEO.

“I didn’t see my mom at work,” Zimmerman says. “I saw Cheryl.”

Unlike sonars that detect fish, FarSounder’s 3D FLS technology was designed to help owners wend past icebergs and coral heads and dodge whales, all while creating their own high-resolution seafloor charts. FarSounder landed its first sale in 2004 and shipped its first product in 2005. “Our first customers were cruise ships and large yachts,” Zimmerman says. The company was also awarded several small-business innovation research grants that enabled years of research and development, and it earned its first (of eight) patents in 2006.

“In 2008 and 2009, we transitioned from R&D work to being a commercial company,” Zimmerman says. “We had around 15 employees at our largest.” This includes Matthew Coolidge, director of hardware development, and Evan Lapisky, director of software engineering, who have both worked at FarSounder since 2002. Zimmerman met them in the university’s jazz band. “Music is acoustics,” he says.

The pandemic brought FarSounder’s first major business headwinds, as supply chains became sticky. The company then moved to semi-remote operations and reconfigured some designs, which got it through the crux.

Listen: Virtual Q&A: FarSounder Argos 350 Forward-Looking Sonar

Things began to settle down, and Matthew Zimmerman took over as CEO in May 2022, a few months after Russia invaded Ukraine. Successive waves of sanctions were imposed, and the company’s second serious business challenge in two years arrived. Oligarchs, after all, adore their superyachts.

Today, the company is still plowing forward. I also spent time with Zimmerman at Wickford Shipyard’s marina. We carried heavy cases down to Cap’n Bert, a 53-foot research vessel owned and operated by the university. Our first stop was the bridge, where we met Capt. Stephen Barber. We were joined by Lapisky and Heath Henley, FarSounder’s senior application engineer (and a guitar player).

The FarSounder team unpacked an Argos 500 (see sidebar) and pole-mounted its transceiver onto Cap’n Bert’s bow. Then we headed out toward Narragansett Bay and the Jamestown Verrazzano Bridge. I had a great view of a laptop running a split-screen view, with two-thirds of the monitor displaying FarSounder’s 3D FLS imagery, and the other third displaying top-down FLS imagery and automatic identification system data layered atop National Oceanic and Atmospheric Administration cartography.

Impressively, the Argos 500 also creates and stores a high-resolution local history map of the ground covered. “When you see something that doesn’t correlate with the chart, that’s what you want to pay attention to,” Zimmerman says.

The local history map looked exactly like how I imagine the seafloor would appear to a scuba diver. As we approached the Jamestown Bridge, I stared at the screen. The working pilings were visible, as were a set of parallel footings. “That’s the old Jamestown Bridge,” Zimmerman says. This latter span was demolished in 2006, and the Argos 500 painted a detailed view of its remnants in customizable colors.

Ahead, a fast ride ripped a white streak through Narragansett Bay’s blue waters. Zimmerman asked Barber to head toward the wake. As we approached, the Argos 500’s range decreased to a few boatlengths.

“Air bubbles are really good acoustic reflectors,” Zimmerman says. “They block the acoustic energy from going to the other side.”

As we cleared the wake, the system’s normal range resumed.

“We’re a software company that makes really big dongles,” Zimmerman told me later, back at the company’s headquarters. “Software makes it possible, but hardware makes it feasible.”

Zimmerman led me to FarSounder’s testing lab and nearby assembly room. “We do all assembly and testing in-house,” he says, adding that FarSounder uses off-the-shelf components whenever possible. Other parts, including the transducers’ piezoelectric ceramics, are manufactured to FarSounder’s specs by third parties—often between larger-volume jobs for other clients. “All components are made in the USA,” he says. “This helps us control quality.” Vendors often warehouse completed components, allowing FarSounder to practice just-in-time manufacturing.

A testing room has a large water tank with a submerged calibrated hydrophone. A hoist lowers FarSounder transceivers into the water, and the hydrophone broadcasts a known frequency sequence to the transducer. Zimmerman points to a monitor that displays the results from 3,600 angles tested simultaneously. “We correct for variance on an individual level,” he says.

Listening to Zimmerman talk about acoustics testing is a reminder that, while he’s mastered 3D FLS sensors, music is his native language.

Coolidge, who designs FarSounder’s electrical components, says the company has made a lot of upgrades to reduce assembly time since the pandemic-era slowdowns. These changes also added future-proofing, but even still, navigating past the Russia sanctions required different thinking. Prior to 2022, Russian ownership accounted for roughly 20 percent to 50 percent of the world’s largest superyachts. Once the sanctions hit, “everything stopped,” Zimmerman says.

The workaround involved adjusting FarSounder’s sales strategy. One green shoot has been the unmanned-surface-vessel market. Another growth trend has been toward yachts with smaller waterlines. Zimmerman hints at a possible smaller system for trawlers; it could be a boon for yacht owners and the scientific community.

In 2023, FarSounder also partnered with Seabed 2030 (see Yachting, May 2024), which aims to map the world’s oceans by 2030. Zimmerman led me to a meeting room where a large screen displayed a FarSounder customer’s recent cruise. “Most of our customers are going places that aren’t well-mapped,” he says, noting that FarSounder sends some clients USB hard drives to capture their systems’ raw data. If issues arise, customers can send the drive to FarSounder, where engineers can troubleshoot and, if necessary, refine the company’s algorithms. These customers can also opt into a fleet-sharing arrangement, where FarSounder sends their anonymous, low-resolution data to Seabed 2030.

In exchange, FarSounder gives these customers access to high-resolution files from the greater fleet-sharing community. This means the customers enjoy some of the world’s finest charts.

And FarSounder sometimes informally collaborates with NOAA scientists. On a recent whale-sounding trip to the Stellwagen Bank National Marine Sanctuary, FarSounder equipment detected humpback whales exhibiting interesting diving behaviors. These findings, Zimmerman says, surprised the NOAA marine biologists.

Zimmerman’s eyes lit up as he talked about FarSounder sensors helping to advance science and protect whales. Listening to him talk, I understood what he meant about bass players and obstacle avoidance: FarSounder’s frequencies may be inaudible, but they’re providing game-changing returns on several levels. 

Forward-Looking Returns

The Argos 350 ($57,000) searches 1,148 feet in front of a vessel’s bow at up to 18 knots. The Argos 500 ($108,000) probes 1,640 feet at 20 knots, and the Argos 1000 ($184,000) can prod 3,281 feet at 25 knots. Each one broadcasts shorter, quieter transmissions for infield detection and longer, louder pings for outfield work.

Take the next step: farsounder.com

The post FarSounder’s Argos Collision-Avoidance System Reviewed appeared first on Yachting.

Sea.AI, a company that uses artificial intelligence technology to try and improve safety at sea through better situational awareness at the helm, premiered its Watchkeeper system this month at the Palm Beach International Boat Show.

Watchkeeper uses Sea.AI’s database of millions of annotated marine objects—floating debris, rafts, buoys, boats not equipped with AIS, and more—to help skippers and crew see hazards and avoid accidents. The idea behind Watchkeeper is to bring this kind of affordable, AI-powered collision-avoidance technology to more recreational boaters, including those who own powerboats and fishing boats of most sizes.

“Our goal with Watchkeeper is to bring advanced safety technology to more boaters, at a price point that makes sense for a broad range of applications,” Marcus Warrelmann, CEO of Sea.AI, stated in a press release. “Collisions are the first event in more than half of all boating accidents and injuries. With automated real-time alerts, Watchkeeper is your extra set of eyes on the water, working in all light conditions and sea states.”

The price point for Watchkeeper starts at $4,999. That includes a 4K low-light camera with an ultrawide field of view; built-in GPS; and Sea.AI software for object recognition. For boaters who need full night-vision capabilities, there’s also a version of the system with integrated long-wave infrared thermal cameras.

Watchkeeper adds to Sea.AI’s existing portfolio of products, which includes the flagship offering called Sentry for large yachts and commercial vessels. Sentry is a 360-degree, AI-powered perimeter surveillance system that provides real-time tracking of people who go overboard, floating hazards, unlit objects and unauthorized approaches, even in total darkness.

Sea.AI also makes products specifically for bluewater sailors, as well as products for racing sailors whose boats have rotating masts. The company’s goal with all of its products is to combine the latest camera technology with artificial intelligence, to give skippers and crewmembers better situational awareness and detection capability than conventional systems such as radar and AIS.

What else did Sea.AI display at the Palm Beach International Boat Show? The company also used the event to promote Brain, which is an add-on product that lets vessels upgrade existing thermal cameras to integrate AI-powered object detection and collision avoidance with automatic alarms. Brain detects small objects, including people who fall overboard, and has real-time alerts to improve situational awareness at the helm.

Take the next step: learn more about all this AI-augmented technology at sea.ai

The post Sea.AI Debuts Watchkeeper appeared first on Yachting.

Boston Harbor is notorious for serving chowder-thick fog when the temperature and dew point align. Such was the case late one night for Brian Moseley and Paul Sullivan, who were enveloped while cruising home aboard Pasithea, their 36-foot Jeanneau NC 1095 Coupe. Fortunately, Pasithea carries a Lookout AI-based collision-avoidance system and an infrared-enabled camera. Instead of staring at the darkened inside of a meteorological pingpong ball, the boaters could use Lookout to see augmented-reality screen views of cans, nuns and nearby shorelines.

Sullivan says the technology allowed them to move along, confident of the location of nearby aids to navigation, islands and the area’s ubiquitous lobster pots. Moseley adds: “I would have felt anxious at the helm without it.”

AI-based watchkeeping and collision-avoidance systems are one of the more exciting pieces of contemporary electronics. While Lookout doesn’t autonomously dodge detected targets or navigational hazards, it does combine AI, a camera, the automatic identification system and cartographic data to provide up to 360-degree situational awareness with collision-avoidance alerts.

Lookout—which has a lower price point than its competition—consists of a marinized black-box processor that connects to a camera, the vessel’s NMEA 2000 data backbone, a 12/24-volt DC power supply and the yacht’s multifunction display.

Owners have a choice of two processors. The Brain ($5,000) is aimed at smaller power yachts and sailboats. It has an Nvidia graphics processing unit that tackles 10 frames per second from a forward-facing high-definition video stream (760 pixels), an augmented-reality processor, an N2K port (for accessing vessel heading and GPS/GNSS information, and for networking with N2K-enabled multifunction displays) and an AIS receiver. It draws 25 watts of continuous DC power with a 60-watt maximum draw.

The Brain Pro ($10,000) is aimed at larger yachts and go-fast rides. It sports similar capabilities, but its Nvidia GPU processes 30 frames per second from multiple full-HD 1080p video streams, providing 360-degree imagery with displayed vessel tracks and buoy annotations. The Brain Pro comes bundled in a larger black box than the Brain and draws 60 watts of continuous power with a 160-watt maximum draw.

Yacht owners can network the Lookout Brain processor with an existing IP-enabled camera—including FLIR thermal-imaging cameras—or spec the Lookout camera ($4,000). While the former can be a great option, the Lookout camera has a long-range HD video feed and includes a near-infrared sensor for nocturnal operations, plus a 360-degree camera for docking (including automotive-style guide lines) and all-around situational awareness.

David Rose, Lookout’s CEO, says that while the system’s hardware is solid, its sorcery resides in its software and implementation of a kind of AI called computer vision. Lookout incorporates three types of computer vision to perceive, identify and track potential threats. This starts with scene segmentation, where the system classifies every pixel in a scene (read: water and not water) and works to stabilize the horizon. A multi-object tracker then follows 100-plus targets within a scene, while a distance-estimation algorithm calculates the range to each target.

Collectively, these AI capabilities allow Lookout to present augmented-reality views on a head-up display on any networked MFD or IP-enabled device, including phones and tablets. It also can present a 3D synthetic view with a bird’s-eye view around the boat. In both cases, chart data is used to create the augmented imagery.

Unlike other AI-based threat-detection systems that consider range using the currency of distance, Rose says, Lookout considers human reaction time and focuses on the potential hazards that lurk in the next 30 seconds. “Radar can detect things miles away, but we focus on extending human perception,” he says. “Thirty seconds at 20 knots is three football fields. We want to be really good at detecting things 30 seconds out.”

Once a target has been detected, Lookout tracks it and uses computer vision to identify it using a deep-learning network that’s been trained and tuned specifically for boating. Rose says Lookout’s algorithm has been trained on hundreds of thousands of images and videos to enable fast, accurate identification. Every Lookout system, he says, captures examples of ambiguous targets and uploads them to Lookout’s cloud, so the system’s model is retrained and pushed out to all Lookout systems.

“We’re using clustering and self-supervised learning,” Rose says, explaining that this largely obviates the need for human labeling.

Lookout systems also use AIS data to detect and identify targets, which the system tracks. Rose says Lookout might someday use MARPA (mini automatic radar plotting aid) data for detecting and tracking distant targets.

On the user-interface side, Lookout provides three alert levels. All detected targets are marked with a white triangle on the augmented-reality view, while targets in the vessel’s path are marked with yellow triangles. Targets that present a collision hazard are marked with a pulsing, automotive-style hazard icon. “Most people like visual alarms more than audible ones,” Rose says.

If a Lookout system “sees” a log or any other navigational danger, it can track the target with a GPS pin and a virtual perimeter. If cloud connectivity exists, it can also share this information with Lookout’s community cloud, which is shared with other connected Lookout users. While Lookout doesn’t require full-time connectivity, cloud access enhances its capabilities and enables over-the-air updates.

In addition to collision avoidance, Lookout can create suggested safe routes using its camera to detect buoys and aids to navigation, and by interrogating the vessel’s vector cartography. Rose says Lookout has also incorporated “Hogwarts-style” graphics, such as hoops over channel entrances and banners that mark a boat owner’s home dock.

While Lookout has been tested aboard vessels ranging from small boats to a 460-foot superyacht, its target market is 30- to 65-footers. Lookout can be added as an aftermarket feature, but Rose says he hopes the technology will eventually be akin to automotive rearview backup cameras, which are standard issue on most new cars. “Tools like this can expand the market and democratize boating,” Rose says.

Lookout’s upsides are quite compelling, but some downsides exist as well. For example, users need to spec out the Brain Pro and the Lookout camera to take advantage of the system’s full capabilities. Also, Lookout can’t use radar-generated MARPA data or autonomously command the vessel’s autopilot to dodge danger.

But given the human tendency to tire during those long watchkeeping duties, there’s little question that Lookout could be a useful helm companion, especially if a chowder-thick fog bank arrives at the wrong time.  

Sensitivity Scale

Lookout allows boaters to adjust the system’s sensitivity level (think radar gain) on a scale of 1 to 5 based on personal preferences. Lower sensitivity equates to greater system confidence that the target exists. Higher sensitivity is likely to catch targets faster, albeit with a greater chance of false positives.

The post Lookout AI-Based Collision Avoidance appeared first on Yachting.

Raymarine has announced a collaboration with Watchit that will see anti-collision technology combined with top-of-the-line chartplotters to help boaters improve situational awareness through display systems at the helm.

In a separate agreement, the companies announced that Raymarine LightHouse Charts are now designated as the official map supplier for Watchit, a change that is expected to enhance the Watchit system’s accuracy and reliability.

Watchit uses algorithms and sensors in a way that’s inspired by automotive safety technology. It analyzes a boat’s data to provide warnings before a collision occurs, helping to avoid collisions as well as groundings.  

“At Raymarine, we’re committed to making boaters’ time on the water safer and more enjoyable,” Michelle Hildyard, vice president of operations at Raymarine, stated in a press release. “Our collaboration with Watchit aligns perfectly with our mission to deliver seamless and open integration with new products and technology, ultimately simplifying the boating experience for our Axiom users.”

The announcement comes at a time when systems integration continues to be a major trend in the recreational marine industry, and as technology continues to be used in new ways to help make boating easier and safer for the skipper and guests. 

When will systems be available with the Watchit-Raymarine integration? They’re available now for compatible Raymarine Axiom chartplotters.

The post Raymarine’s Upgraded Collision Control appeared first on Yachting.

Some lessons must be learned the hard way. Take the Atlantis, an 80-foot express cruiser that was motoring 3 miles off the coast of St. Augustine, Florida, this past May during a Memorial Day weekend voyage. The yacht was reportedly operating under mostly clear skies when it struck an object, likely a large metal marker denoting a submerged dredge pipe. At 11:37 a.m., the US Coast Guard in Jacksonville received emergency notification via VHF channel 16 that Atlantis was sinking. The Coast Guard dispatched a boat and contacted the St. Johns County Fire Rescue Division, which rescued two mariners.

While it’s unclear how Atlantis’ crew failed to spot the marker, better watchkeeping and collision-avoidance technology likely could have prevented this accident.

Watchkeeping often involves long periods of monotony punctuated by occasional moments of stress. AI-enabled technologies can ease this burden. Sea Machines’ AI-ris (pronounced “eye-ris”) Computer Vision Sensor alerts recreational mariners about dangerous targets. Although it doesn’t autonomously evade danger, the system processes targets in milliseconds, supports fast cruising speeds and enhances situational awareness.

AI-ris ($27,900) is an optical-based system that uses AI technology called computer vision to detect, classify, geolocate and track multiple targets. The system accomplishes this via a custom machine-learning model that Sea Machines trained on millions of images. “AI-ris is designed to enhance situational awareness for all vessels under power that are 33 feet or longer,” says James Miller, Sea Machines’ AI-ris product manager.

The system has a forward-looking daylight camera and a rugged black-box processor. It also has a touchscreen user-interface screen, or boaters can substitute a compatible Furuno or Raymarine multifunction display, or a generic touchscreen display. AI-ris requires NMEA 2000 connectivity to access the vessel’s GPS/GNSS sensor, and it can be spec’d with a Sea Machines thermal-imaging camera.

The daylight camera must be mounted at least 25 feet above the waterline to deliver its full range of 5 nautical miles. AI-ris can simultaneously classify and track 50 targets. In terms of target size, AI-ris can detect a 13-foot object at 0.25 nautical miles; at 1 mile, it can detect a 49- to -59-foot object; and at 5 miles, it can detect a 246- to 295-foot object. AI-ris reportedly has 99 to 100 percent accuracy when it can place 20 pixels on a target.

Given that this target classification wizardry resides in a machine-learning model, the optical-based system requires imagery. The daylight camera captures 30 frames per second, has a 90-degree field of view (horizontal and vertical), and uses a low-light mode to capture moon- and starlight. The 10-megapixel sensor yields 4K onscreen imagery that’s shared with the processor via an Ethernet cable.

Processed imagery is then streamed onto the user’s screen. Users can take screen grabs and capture video, a feature that Miller says can be useful in documenting incidents.

“The custom machine-learning model was trained on over 25 million images of vessels and objects [taken] from a variety of vessels operating globally in different sea and lighting conditions,” Miller says.

These images have yielded more than 35 million examples of marine targets—but the system doesn’t work like a search engine. “Rather than looking up a vessel or object within a database, the computer-vision model recognizes important objects in view by its understanding of how these objects appear and behave,” he says.

AI-ris does this very quickly. Miller says it will detect, classify and track multiple targets in less than 250 milliseconds. Depending on environmental factors, the number of targets and the distance to the object, he adds, “This can occur in significantly less time.”

On the user-interface side, AI-ris creates a 2D augmented-reality display on its networked screen. Targets are graphically boxed, color-coded and placed into four classification buckets. Yellow indicates powerboats, blue denotes sailboats, and green represents marine mammals. White refers to miscellaneous objects, including aids to navigation, kayaks, swimmers and logs. Miller says Sea Machines is adding eight additional classification buckets soon.

Alternatively, AI-ris can display a radar-style target-range viewer that depicts the vessel in the center, with outward-extending range bands. Targets appear as color-coded triangles, providing classification along with visual range and bearing information.

Users can set guard zones (think radar) in both modes. Once a target is detected, classified and tracked within a guard zone, AI-ris provides visual and auditory warnings. These begin with a banner at the top of the screen; optional auditory alerts are played twice, 20 seconds apart, while the escalatory auditory alarms are played every 20 seconds until the alarms are manually cleared.

“Customer feedback emphasized that the system shouldn’t become something that an operator has to constantly attend to; rather, [it’s] something that supports safe navigation,” Miller says. “For this reason, we have concentrated on a fine balance between passive and active notifications.”

While AI-ris has interesting capabilities, it has limitations like all technologies. For example, its 25-foot mounting-height requirement is a big ask for smaller yachts. The camera’s 90-degree field of view leaves a 270-degree blind spot unless the vessel also carries an automatic identification system receiver or radar. The system isn’t compatible with third-party cameras, and it can’t draw information from the vessel’s vector cartography to verify the position of, say, aids to navigation. As of this writing, the system also can’t autonomously command the autopilot to avoid collisions.

That said, AI-ris does provide unflagging situational awareness within its field of view. It supports vessel speeds up to 45 knots, and technology is on the way that will provide the same daylight-camera functionality on the company-supplied thermal-imaging cameras. Sea Machines is planning a release that will allow a yacht with AI-ris, a Sea Machines SM360 advanced autopilot system, and a Rolls-Royce power and bridge system to dodge dangerous targets autonomously.

For now, if the goal is to enjoy less-memorable Memorial Day weekends than the crew on Atlantis experienced, the Sea Machines AI-ris system provides a tireless eye on the horizon.

Lifelong Learners

AI-ris is based on a model that has already been trained on more than 25 million images, but more data equals increased safety. Sea Machines collects new imagery using a fleet of test boats and working with customers who share voyage data. The company then releases yearly updates that expand the model’s identification capabilities.

The post Introducing AI-ris from Sea Machines appeared first on Yachting.