| Sonstiges: |
- Nachgewiesen in: USPTO Patent Applications
- Sprachen: English
- Document Number: 20240051155
- Publication Date: February 15, 2024
- Appl. No: 18/486936
- Application Filed: October 13, 2023
- Claim: 1. A method, comprising: obtaining inspection path information indicating a path for a robot to travel, and a plurality of locations along the path to inspect, wherein each location of the plurality of locations is associated with orientation information indicating an orientation that an OGI camera coupled to the robot is to be placed in to record images; causing the robot to move along the path; determining that the robot is at a first location of the plurality of locations; causing the robot to adjust the OGI camera based on first orientation information associated with the first location; recording a first plurality of images with the OGI camera; determining, based on the first plurality of images, that a gas leak exists at the first location, wherein determining that the gas leak exists at the first location comprises: determining a reference image associated with the first location; generating a subtracted image by subtracting a first image of the first plurality of images from the reference image; and determining, based on the subtracted image, that the gas leak exists at the first location; and storing, in memory, an indication that the gas leak exists at the first location.
- Claim: 2. The method claim 1, comprising: in response to determining that the gas leak exists at the first location, recording, via an RGB camera, an image of the first location; and sending the image to a server.
- Claim: 3. The method claim 1, wherein determining that the gas leak exists at the first location comprises: classifying, via a convolutional neural network (CNN), whether the first image depicts the gas leak.
- Claim: 4. The method claim 1, wherein determining a reference image comprises: obtaining, from a database, a historical image of the first location, wherein the historical image depicts an absence of the gas leak at the first location on a date that the historical image was taken.
- Claim: 5. The method claim 1, wherein determining a reference image comprises: selecting, from the first plurality of images, an image that was recorded before any other image of the first plurality of images.
- Claim: 6. The method claim 1, wherein determining that the gas leak exists at the first location comprises: generating, based on inputting the subtracted image into a machine learning model, a similarity score; comparing the similarity score with a threshold similarity score; and based on comparing the similarity score with a threshold similarity score, determining that the gas leak exists at the first location.
- Claim: 7. The method claim 1, wherein determining that the gas leak exists at the first location comprises: obtaining a second plurality of historical images associated with the first location; generating, based on inputting the first plurality of images into a machine learning model, a first vector representation of the first plurality of images; generating, based on inputting the second plurality of images into a machine learning model, a second vector representation of the second plurality of historical images; and determining, based on a comparison of the first vector with the second vector, that the gas leak exists at the first location.
- Claim: 8. The method claim 1, comprising: recording, via the OGI camera, a set of images, wherein the set of images comprises an image for each location of the plurality of locations; generating a label for each image in the set of images, wherein each label indicates a location associated with a corresponding image and indicates whether the gas leak was detected in the corresponding image; and sending the set of images to a server for use in training a machine learning model.
- Claim: 9. The method claim 1, comprising: inputting the subtracted image into a machine learning model; and in response to inputting the first plurality of images into the machine learning model, classifying one or more objects in the subtracted image as a plume of gas.
- Claim: 10. The method of claim 9, comprising: determining, based on output from the machine learning model and based on the subtracted image, a target location within the plume of gas; sensing, via a laser sensor, a concentration level of gas at the target location.
- Claim: 11. The method claim 1, comprising: receiving a second plurality of images captured by the OGI camera, wherein the second plurality of images are associated with one or more locations of the plurality of locations; generating a webpage comprising a user interface, wherein the user interface is configured to receive input on one or more portions of an image; receiving, via the webpage, input corresponding to one or more images of the second plurality of images, wherein the input indicates the gas leak exists in the one or more images of the second plurality of images; and generating, based on the input and the second plurality of images, a training data set for a machine learning model.
- Claim: 12. The method claim 1, comprising, wherein the first orientation information comprises a vector indicating a roll of the OGI camera, a pitch of the OGI camera, and a yaw of the OGI camera.
- Claim: 13. The method claim 1, wherein causing the robot to move along the path comprises: receiving information indicating a user that is logged into the robot; determining location permissions associated with the user; and causing, based on the location permissions associated with the user, the robot to skip a second location of the plurality of locations.
- Claim: 14. The method claim 1, comprising causing a server configured to perform operations comprising: receiving the first plurality of images; determining, based on inputting the first plurality of images into a machine learning model, that there is no gas leak at the first location; and in response to determining that there is no gas leak at the first location, sending an indication that no gas leak was detected in the first plurality of images and a request to the robot to record a second plurality of images of the first location.
- Claim: 15. The method of claim 14, comprising: in response to receiving the request at the robot, recording a second plurality of images of the first location; determining, based on the second plurality of images, that there is no gas leak at the first location; and in response to determining that there is no gas leak at the first location, causing the robot to move to a second location of the plurality of locations.
- Claim: 16. The method of claim 14, comprising: in response to receiving the request, causing the robot to move closer to the first location; recording a second plurality of images of the first location; and sending the second plurality of images to the server.
- Claim: 17. The method of claim 1, comprising causing a server system to perform operations comprising: obtaining the inspection path information; obtaining location information corresponding to a plurality of robots, wherein the plurality of robots comprises the robot; determining, based on the location information, that the robot is closer to the first location than other robots of the plurality of robots; and in response to determining that the robot is closer to the first location than other robots of the plurality of robots, sending the inspection path information to the robot.
- Claim: 18. The method of claim 1, wherein the robot is configured to perform operations comprising: determining that more than a threshold amount of time has transpired since receiving a heartbeat message; and in response to determining that more than a threshold amount of time has transpired, sending a request to a server, wherein the request indicates a diagnostic procedure for the server to perform on at least some of a plurality of sensors of the robot.
- Claim: 19. The method of claim 1, wherein the robot is configured to perform operations comprising: determining that more than a threshold amount of time has transpired since receiving a heartbeat message; and in response to determining that more than a threshold amount of time has transpired, moving to a charging station associated with the robot.
- Claim: 20. The method of claim 1, wherein the gas leak is a methane leak from an oil or gas processing facility.
- Claim: 21. The method of claim 1, comprising: steps for training one or more machine learning models to detect gas leaks.
- Claim: 22. The method of claim 1, wherein at least some steps of the method are performed with the robot and at least some of the operations are performed with a remote server.
- Claim: 23. The method of claim 1, wherein detecting the gas leak comprises steps for detecting whether a plume of gas exists in an image taken by the OGI camera.
- Claim: 24. The method of claim 1, wherein detecting the gas leak comprises: determining, in pixel space of the subtracted image, a perimeter of a region including a plume of gas, and wherein.
- Claim: 25. The method of claim 1, wherein: the robot is a quadruped robot; the gas is methane gas; the first plurality of locations are locations of a hydrocarbon processing facility; and the OGI camera comprises a thermoelectric cooler configured to cool an image sensor of the OGI camera.
- Claim: 26. The method of claim 1, wherein: the robot is configured to fly.
- Current International Class: 25; 25; 25; 06; 06; 06; 05
|
