Aerial monitoring system and method for identifying and locating object features

A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to:

• receive training images of training objects, wherein the training images include predefined object features indicating a physical state of the training objects;
• build a convolutional neural network using the training images;
• receive monitoring images of monitoring objects in a common area and geotags indicating global positioning system coordinates of the monitoring objects appearing in the monitoring images;
• mosaic the monitoring images to create a single mosaicked image, wherein the single mosaicked image includes monitoring image object features each indicating a physical state of the monitoring objects;
• use the convolutional neural network to analyze the monitoring image object features for similarity with the predefined object features;
• use the convolutional neural network to categorize the monitoring images into maintenance issue categories based on the physical state of the monitoring images indicated by the analyzed monitoring image object features;
• use the global positioning system coordinates of the monitoring objects together with a fuzzy clustering process to group the monitoring objects into a plurality of cluster locations based on proximity between monitoring objects, each cluster location defining a region within the common area, wherein each cluster location includes a center and more than one monitoring object of the monitoring objects;
• determine which cluster location of the plurality of cluster locations each monitoring image object feature belongs to;
• generate a map displaying map images of the monitoring objects in positions corresponding to their global positioning system coordinates and display indicia identifying the plurality of cluster locations;
• and
• output and display a first maintenance issue category and a first cluster location for a first monitoring object of the monitoring objects based on the results of both categorizing the monitoring images and determining which cluster location of the plurality of cluster locations each monitoring image object feature belongs to.

An aerial monitoring system for identifying and locating object features to create a maintenance plan, comprising:

• one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to:
• receive training images of training objects, wherein the training images include predefined object features indicating a physical state of the training objects;
• build a convolutional neural network using the training images;
• receive monitoring images of monitoring objects in a common area and geotags indicating global positioning system coordinates of the monitoring objects appearing in the monitoring images;
• mosaic the monitoring images to create a single mosaicked image, wherein the single mosaicked image includes monitoring image object features each indicating a physical state of the monitoring objects;
• use the convolutional neural network to analyze the monitoring image object features for similarity with the predefined object features;
• use the convolutional neural network to categorize the monitoring images into maintenance issue categories based on the physical state of the monitoring images indicated by the analyzed monitoring image object features;
• use the global positioning system coordinates of the monitoring objects together with a fuzzy clustering process to group the monitoring objects into a plurality of cluster locations based on proximity between monitoring objects, each cluster location defining a region within the common area, wherein each cluster location includes a center and more than one monitoring object of the monitoring objects;
• determine which cluster location of the plurality of cluster locations each monitoring image object feature belongs to;
• generate a map display map images of the monitoring objects in positions corresponding to their global positioning system coordinates and display indicia identifying the plurality of cluster locations;
• and
• output and display a first maintenance issue category and a first cluster location for a first monitoring object of the monitoring objects based on the results of both categorizing the monitoring images and determining which cluster location of the plurality of cluster locations each monitoring image object feature belongs to.