Cloud detection on remote sensing imagery

A non-transitory computer-readable storage medium storing one or more instructions which, when executed by one or more processors, cause the one or more processors to perform steps comprising:

• receiving remote sensing imagery of a geographical area, wherein the remote sensing imagery includes a plurality of pixels and one or more band values for each pixel within the remote sensing imagery;
• using a first classifier, identifying one or more cloud seed pixels from the remote sensing imagery based on the one or more band values for each pixel;
• using a second classifier, identifying one or more candidate cloud pixels from the remote sensing imagery based on the one or more band values for each pixel, wherein the first classifier favors precision compared to the second classifier and the second classifier favors recall compared to the first classifier;
• identifying a set of the one or more candidate cloud pixels that are connected to the one or more cloud seed pixels by one or more of:
• being a member of the one or more cloud seed pixels, being directly connected to at least one of the one or more cloud seed pixels, or being indirectly connected to at least one of the one or more cloud seed pixels through at least one candidate cloud pixel of the one or more candidate cloud pixels;
• generating a cloud mask for the remote sensing imagery based on the set, wherein the cloud mask identifies which pixels within the remote sensing imagery correspond to clouds.

A data processing system comprising:

• a memory;
• one or more processors coupled to the memory;
• cloud seed generator logic stored in the memory, executable by the one or more processors, and configured to cause the one or more processors to:
• receive remote sensing imagery of a geographical area, wherein the remote sensing imagery includes a plurality of pixels and one or more band values for each pixel within the remote sensing imagery,
• use a first classifier to identify one or more cloud seed pixels from the remote sensing imagery based on the one or more band values for each pixel;
• cloud mask generator logic stored in the memory, executable by the one or more processors, and configured to cause the one or more processors to:
• use a second classifier to identify one or more candidate cloud pixels from the remote sensing imagery based on the one or more band values for each pixel, wherein the first classifier favors precision compared to the second classifier and the second classifier favors recall compared to the first classifier,
• identify a set of the one or more candidate cloud pixels that are connected to the one or more cloud seed pixels by one or more of:
• being a member of the one or more cloud seed pixels, being directly connected to at least one of the one or more cloud seed pixels, or being indirectly connected to at least one of the one or more cloud seed pixels through at least one candidate cloud pixel of the one or more candidate cloud pixels,
• generate a cloud mask for the remote sensing imagery based on the set, wherein the cloud mask identifies which pixels within the remote sensing imagery correspond to clouds.