If you aren't sure where to begin, one way to get your feet wet with this project is to start by co-registering a set of images. There are ways to do this programmatically, but many methods are patented and built into tools like ESRI. However, the most accurate way to co-register two images (where one has no coordinate system) is to use control points.
\n\nI haven't found a good open source tool to automatically select ground control points, but you can do it yourself in QGIS. Here's a tutorial on how to set it up: http://www.qgistutorials.com/en/docs/georeferencing_basics.html
\n\nYou can either co-register images to each other (based on X/Y) or, if you know where they are in California, to an actual coordinate system. If you work manually, remember to select control points on the ground, not on rooftops. The QGIS georeferencer won't accurately orthorectify (and you probably won't have any elevation data more accurate than DTED, if you even have that) so don't put points on hills or your image will skew.
\n\nAfter you have co-registered a set of images, it is much easier to work on change analysis. Then if you develop a system for change analysis, you can find a more generalized way to co-register images.
\n\nHappy to field any questions about georeference ground control points or hear about Auto GCP tools that folks have found helpful.
", "messages": [ { "id": "120048", "postDate": "05/14/2016 22:58:44", "content": "If you aren't sure where to begin, one way to get your feet wet with this project is to start by co-registering a set of images. There are ways to do this programmatically, but many methods are patented and built into tools like ESRI. However, the most accurate way to co-register two images (where one has no coordinate system) is to use control points.
\n\nI haven't found a good open source tool to automatically select ground control points, but you can do it yourself in QGIS. Here's a tutorial on how to set it up: http://www.qgistutorials.com/en/docs/georeferencing_basics.html
\n\nYou can either co-register images to each other (based on X/Y) or, if you know where they are in California, to an actual coordinate system. If you work manually, remember to select control points on the ground, not on rooftops. The QGIS georeferencer won't accurately orthorectify (and you probably won't have any elevation data more accurate than DTED, if you even have that) so don't put points on hills or your image will skew.
\n\nAfter you have co-registered a set of images, it is much easier to work on change analysis. Then if you develop a system for change analysis, you can find a more generalized way to co-register images.
\n\nHappy to field any questions about georeference ground control points or hear about Auto GCP tools that folks have found helpful.
", "rawMarkdown": "If you aren't sure where to begin, one way to get your feet wet with this project is to start by co-registering a set of images. There are ways to do this programmatically, but many methods are patented and built into tools like ESRI. However, the most accurate way to co-register two images (where one has no coordinate system) is to use control points. \r\n\r\nI haven't found a good open source tool to automatically select ground control points, but you can do it yourself in QGIS. Here's a tutorial on how to set it up: http://www.qgistutorials.com/en/docs/georeferencing_basics.html\r\n\r\nYou can either co-register images to each other (based on X/Y) or, if you know where they are in California, to an actual coordinate system. If you work manually, remember to select control points on the ground, not on rooftops. The QGIS georeferencer won't accurately orthorectify (and you probably won't have any elevation data more accurate than DTED, if you even have that) so don't put points on hills or your image will skew. \r\n\r\nAfter you have co-registered a set of images, it is much easier to work on change analysis. Then if you develop a system for change analysis, you can find a more generalized way to co-register images. \r\n\r\nHappy to field any questions about georeference ground control points or hear about Auto GCP tools that folks have found helpful.", "votes": null }, { "id": "120064", "postDate": "05/15/2016 03:02:45", "content": "Thanks a lot for the information. I will try that soon.
", "rawMarkdown": "Thanks a lot for the information. I will try that soon.", "votes": null } ], "comments": [ { "id": 120064, "author_name": "yourwanghao", "author_url": "", "post_date": "05/15/2016 03:02:45", "content": "Thanks a lot for the information. I will try that soon.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "120048": "If you aren't sure where to begin, one way to get your feet wet with this project is to start by co-registering a set of images. There are ways to do this programmatically, but many methods are patented and built into tools like ESRI. However, the most accurate way to co-register two images (where one has no coordinate system) is to use control points. \r\n\r\nI haven't found a good open source tool to automatically select ground control points, but you can do it yourself in QGIS. Here's a tutorial on how to set it up: http://www.qgistutorials.com/en/docs/georeferencing_basics.html\r\n\r\nYou can either co-register images to each other (based on X/Y) or, if you know where they are in California, to an actual coordinate system. If you work manually, remember to select control points on the ground, not on rooftops. The QGIS georeferencer won't accurately orthorectify (and you probably won't have any elevation data more accurate than DTED, if you even have that) so don't put points on hills or your image will skew. \r\n\r\nAfter you have co-registered a set of images, it is much easier to work on change analysis. Then if you develop a system for change analysis, you can find a more generalized way to co-register images. \r\n\r\nHappy to field any questions about georeference ground control points or hear about Auto GCP tools that folks have found helpful.", "120064": "Thanks a lot for the information. I will try that soon." }, "source": "meta" }