a a a a a a a a a

MaritimeAI

AI in maritime* industry and this is epic.

Our mission

The Problem

Maritime resources are believed to be our second hope when our earth ones are depleted. Indeed, ocean represents 76% of the world trade an 43% of undiscovered oil&gas deposits are believed to be under the sea bed. Yet the way those resources are being explored, used and controlled is not only only largely ineffective but also harmful and dangerous for the future of our planet and population.

The mission

At MaritimeAI we aim to introduce edge technologies to increase the effectiveness of resources control, by decreasing both costs and danger for the ecosystem of the Earth.

The technology

We apply big data and AI tools and frameworks in order to create a sustainable maritime economy for all major and minor actors, offering both box and custom solution scalable from small fisheries to huge multinational oil&gas giants. Our way of working is agile at heart and business oriented in in mind, disrupting the way the things are so they become the way they should be in order to have strong, healthy and innovative maritime economy of the future.

Core members

Pavel Golubev

CEO, Strategy, Technology

Andrew Kohan, PhD

CSO, Research, Science




Our value propositions

Ice detection, classification and analytics on imagery







Problem statement

Development of shipping and marine constructions in Arctic requires precise and operative data on ice conditions and their historical change. Different imagery sources each have benefits and disadvantages and standalone decision is unfit for problem solving. Now there is need in complex decision combining imagery flow from various sources (SAR, multispectral and aerial imagery, ship observations) and as fast as possible imagery analysis.

Unique solution

We utilize neural networks for detection, classification and analytics of ice conditions including ice drift and deformation prediction. Our service is able to work with Sentinel-1, 2, Landsat-8, drone imagery and provides needed data in minutes otherwise of days.

Customer benefit

Customers have access to very fast and high-resolution ice analytics and are able to perform decision-making operatively optimizing routing and ice protection measures. It reduces costs for fuel, freight, C02 emissions and faster cargo delivery.

Example

Multiclass segmentation

Area covered by test image is approximately 122.8×122.8 km ~15000 km² Resolution of image is fully preserved and each pixel of image 4096x4096 pixels in size is classified.

  • Open water
  • Fractures in old ice area
  • Water with low concentration of new ice
  • Old ice
  • New ice
  • Young ice

Automated seabed segmentation and object detection on side scan sonar imagery

Problem statement:

Worldwide development of offshore activities including oil and gas projects, pipeline and cable laying activities increases volume of seabed sonar imagery data for pre-construction survey, inspection and decommissioning of subsea assets. Distinguishing objects on seabed and delineating of different seabed types requires labour intensive data interpretation with very high requirements to interpreter attentiveness and experience. The other issue is ambiguity of interpretation and inevitability of mistakes due to human factor

Unique solution:

we utilize neural networks for detection and classification of objects on seabed and seabed types segmentation. Having as input processed xtf or other common sonar data formats our service provides catalogue of seabed objects in csv format with measurements of objects, objects snapshots and polygons of seabed types. It requires input of 5-10% of manually interpreted data and supports additional information of underwater sampling, imaging, acoustic profiling and bathymetry interpretation.

Customer benefit:

Customers have access to very fast automatic side scan sonar imagery interpretation tool which provides data for subsequent eventing, environmental and geological analyses of the survey area thus decreasing time and money spending and leaving more time for deeper data analyses and reporting.

Example

Predicted

Ground truth

Automated seismic interpretation

Problem statement

Worldwide development of oil and gas projects in increasingly difficult geological conditions (for example, deepwater offshore projects) leads to increase of seismic data volumes and has-text-light requires more labour-intensive data interpretation. Also in such circumstances there is increase in fast and unambiguous data interpretation. In addition, in the conditions of tight deadlines and / or large volumes, the interpreter inevitably makes mistakes, for example, misses various features of the geological section.

Unique solution

We utilize neural networks for detection and classification of geological layers, reflecting horizons, seismic facies. Having as input 2D (horizontal or vertical seismic sections) or 3D (seismic cubes) data our service provides fast detection of very time-consuming and ambiguously interpreted geological features such as salt bodies, acoustic anomalies and faults. Given as input of manually labeled less than 5% of survey data service outputs automatically interpreted data in minutes instead of days. Output is possible in any common and specialized formats for further analytics in seismic interpretation software

Customer benefit

Customers have access to very fast automatic seismic interpretation tool which provides very fast insights into geological structure of the certain area. It enables easier and faster workflow for further survey planning, data interpretation correction and reservoir modeling thus decreasing time and money spending and leaving more time for deeper data analyses.

Example

Automated object detection on side scan sonar imagery for search and recovery purposes

Problem statement

Worldwide development of marine activities and air traffic over seas and oceans lead to increase of incidents on open seas. Search and recovery operations use side scan sonar data to discover aftermath of these incidents. Now search is performed manually by operators.

Unique solution

we utilize neural networks for detection and classification of man-made objects on seabed. Having as input raw xtf or other common sonar data formats our service provides fast detection of drowned man-made objects

Customer benefit

Customers have access to very fast automatic man-made seabed object detection tool reducing time-to-discovery from days to hours.

Mars

Computer vision can help us not only with detection of SEA ice, but also with detection of SUBSURFACE ice. We started a project targeted on development of computer vision technologies for detection of ice-related surface features on Mars.

This planet is the first target for human colonization in solar system. Future colonists will face need in vital resource - water, needed for both maintaining colonies and production of hydrogen for rocket fuel.

Various sources of satellite imagery provide enough information for detail mapping of ice-related features on Mars, thus giving insights into stockpiles of subsurface ice. The bottleneck is as usual: huge amount of data and shortage of human effort and time. Neural networks and computer vision technology can give excellent solution.

First results of our project showing automated detection of martian ice-wedge polygons are below

Example

Binary segmentation

The images show an example of segmentation of the surface of Mars, the so-called. permafrost "polygons", which, according to geologists, formed under the influence of water ice.

Geo Platform



We offer a universal geoplatform with the following features:

  • Realtime ice segmentation
  • Ice coverage analytics
  • AIS analytics and routing
  • Custom raster and vector layers
  • AI tools
  • Oil spills detection

Our partners



Поиск под водой
Nvidia
Amazon
Azure
Yandex.Cloud
ODS
ЦМИ МГУ
gibmirpresence