Road and infrastructure analysis tool
AssigneeNEC Corporation
InventorsSparsh Garg, Samuel Schulter, Vijay Kumar Baikampady Gopalkrishna
Life of this patent
Estimating the term…
Abstract
Systems and methods is provided for road hazard analysis. The method includes obtaining sensor data of a road environment including a road and observable surroundings, and applying labels to the sensor data. The method further includes training a first neural network model to identify road hazards, training a second neural network model to identify faded lane markings, and training a third neural network model to identify overhanging trees and blocking foliage. The method further includes implementing the trained neural network models to detect road hazards in a real road setting.
Claims20
A method for road hazard analysis, comprising:
- obtaining sensor data of a road environment including a road and observable surroundings;
- transforming, using a universal representation learning model, the sensor data into semantic maps by applying pseudo labels of road scene categories, generated through semantic segmentation of the road scene categories, to a road scene image obtained from the sensor data including:
- training a first neural network model to identify road hazards with the semantic maps;
- training a second neural network model to identify faded lane markings with the semantic maps;
- training a third neural network model to identify overhanging trees and blocking foliage with the semantic maps by determining a percentage of trees over the road based on filtered points from point clouds converted from generated depth maps by a depth model based on the semantic maps;
- generating, with the first, second, and third neural network models, a top-view road map having overlays of identified road hazards and maintenance instructions for the identified road hazards in a real world setting;
- and
- communicating, through communications hardware, the top-view road map to one or more work crews.
A system for road hazard analysis, comprising:
- one or more processors;
- memory coupled to the one or more processors through a bus;
- and
- communication hardware coupled to the one or more processors and memory,
- wherein the memory operates with the one or more processors and the communication hardware to perform:
- obtaining sensor data of a road environment including a road and observable surroundings;
- transforming, using a universal representation learning model, the sensor data into semantic maps by applying pseudo labels of road scene categories, generated through semantic segmentation of the road scene categories, to a road scene image obtained from the sensor data;
- training a first neural network model to identify road hazards with the semantic maps;
- training a second neural network model to identify faded lane markings with the semantic maps;
- training a third neural network model to identify overhanging trees and blocking foliage with the semantic maps by determining a percentage of trees over the road based on filtered points from point clouds converted from generated depth maps by a depth model based on the semantic maps;
- generating, with the first, second, and third neural network models, a top-view road map having overlays of identified road hazards and maintenance instructions for the identified road hazards in a real world setting;
- and
- communicating, through communications hardware, the top-view road map to one or more work crews.
Dependencies are inferred from the claim text.