Broad area geospatial object detection using autogenerated deep learning models

1. A system for broad area geospatial object detection using auto-generated deep learning models, comprising:

a deep learning model training software module stored in a memory of and operating on a processor of a computing device; and

an image analysis software module stored in the memory of and operating on the processor of the computing device;

wherein the deep learning model training software module:

receives training data comprising a plurality of orthorectified geospatial images with a plurality of labeled objects present therein;

classifies the training data into a plurality of categories;

applies one or more image modification steps to the training data; and

generates an object classification model from the training data using a deep learning method comprising processing of the training data through a convolutional neural network system; and

wherein the image analysis software module:

receives orthorectified geospatial imagery;

applies a plurality of image modifications to the orthorectified geospatial imagery;

discards images unsuitable for analysis;

uses the object classification model to automatically identify and label all objects of interest in the received orthorectified geospatial imagery; and

outputs the geospatial locations of the identified objects of interest.

2. A method for conducting automated identification of feature items from orthorectified geospatial imagery data comprising the following steps:

receiving, at a deep learning model training software module stored in a memory of and operating on a processor of a computing device, training data comprising a plurality of orthorectified geospatial images with a plurality of labeled objects present therein;

classifying the training data into a plurality of categories using the deep learning model training software module;

applying one or more image modification steps to the training data using the deep learning model training software module;

generating an object classification model from the training data using a deep learning method comprising processing of the training data through a convolutional neural network system operating as part of the deep learning model training software module;

receiving orthorectified geospatial imagery at an image analysis software module stored in the memory of and operating on the processor of the computing device;

applying a plurality of image modifications to the received orthorectified geospatial imagery using the image analysis software module;

discarding images unsuitable for analysis from the modified orthorectified geospatial imagery using the image analysis software module;

retaining images suitable for analysis from the modified orthorectified geospatial imagery using the image analysis software module;

using the object classification model to automatically identify and label all objects of interest in the retained orthorectified geospatial imagery using the image analysis software module; and

outputting the geospatial locations of the identified objects of interest using the image analysis software module.

3. The system of claim 1 , wherein the plurality of categories are pre-defined.

4. The system of claim 1 , wherein the plurality of image modification steps are drawn from a set comprising artifact removal, color standardization, and resolution data determination.

5. The system of claim 1 , wherein the objects of interest are automatically identified and labeled regardless of the orientation of that feature item within the section.

6. The system of claim 5 , further wherein the objects of interest are automatically identified and labeled accounting for differences in item scale by using a multi-scale sliding window algorithm.

7. The method of claim 2 , wherein the plurality of categories are pre-defined.

8. The method of claim 2 , wherein the plurality of image modification steps are drawn from a set comprising artifact removal, color standardization, and resolution data determination.

9. The method of claim 2 , wherein the feature items of interest are identified regardless of the orientation of that feature item within the section.

10. The method of claim 9 , further wherein the feature items of interest are identified accounting for differences in item scale by using a multi-scale sliding window algorithm.