Dynamic radiometric thermal imaging compensation

A method for dynamic radiometric thermal imaging compensation, the method comprising:

• receiving, from a visible light imager having a first field of view, a visible light image of a scene, the visible light image including a visible light image overlap section;
• analyzing the visible light image using a convolutional neural network to perform semantic image segmentation on the visible light image to assign a material type to each of the plurality of visible light pixels making up the visible light image;
• determining an emissivity value for each of the plurality of visible light pixels making up the visible light image based on the material type assigned to the visible light pixel;
• receiving, from a thermal imager having a second field of view overlapping the first field of view, a thermal image of the scene, the thermal image including a thermal image overlap section that corresponds to the visible light image overlap section, the thermal image including a plurality of thermal pixels, each of the plurality of thermal pixels having a thermal pixel value;
• associating, with the electronic processor, each of the plurality of thermal pixels making up the thermal image overlap section with at least one of the plurality of visible light pixels making up the visible light image overlap section;
• and
• generating, with the electronic processor, a second thermal image by, for each of the plurality of thermal pixels making up the thermal image overlap section, determining a temperature value based on the thermal pixel value of the thermal pixel and the emissivity value of the at least one of the plurality of visible light pixels associated with the thermal pixel.

An unmanned aerial vehicle comprising:

• a thermal imager having a first field of view;
• a visible light imager having a second field of view overlapping the first field of view;
• an electronic processor coupled the thermal imager and the visible light imager and configured to:
• receive, from the visible light imager, a visible light image of a scene, the visible light image including a visible light image overlap section;
• analyze the visible light image by using a convolutional neural network to perform semantic image segmentation on the visible light image to assign a material type to each of the plurality of visible light pixels making up the visible light image;
• determine an emissivity value for each of the plurality of visible light pixels making up the visible light image based on the material type assigned to the visible light pixel;
• receive, from the thermal imager, a thermal image of the scene, the thermal image including a thermal image overlap section that corresponds to the visible light image overlap section, the thermal image including a plurality of thermal pixels, each of the plurality of thermal pixels having a thermal pixel value;
• associate each of the plurality of thermal pixels making up the thermal image overlap section with at least one of the plurality of visible light pixels making up the visible light image overlap section;
• and
• generate a second thermal image by, for each of the plurality of thermal pixels making up the thermal image overlap section, determining a temperature value based on the thermal pixel value of the thermal pixel and the emissivity value of the at least one of the plurality of visible light pixels associated with the thermal pixel.