Scripted selection of reference images from a series of images
Posted: 2019-07-01T16:20:53-07:00
Greetings!
I have thousands of images of cars entering or exiting a parking garage taken by the same surveillance camera. In fact, I have two such series of images taken from two different cameras.
Some of the images have cars in them. Other images in the same series have a picture of the environment without a car. Using the compare function of ImageMagick's suite of apps, is there an industry-accepted method of comparing the pictures in the series by script to other images in the same series to "learn" or predict which images have cars in them? I believe that such an image of the surroundings but lacking the primary interest of the series is called a reference image.
Ultimately, I want to generate a set of reference images for the whole day (one or more per hours) which would allow for the detection of cars by comparison with the reference image by script.
Here is a link to a reference image: creds = lpr / 1q2w3e4r5t.!.
https://lpr.hathcock.org/backdoor/1800W ... NTRACT.jpg
When there is a car in the image, it is quite different. As in this case:
https://lpr.hathcock.org/backdoor/1800W ... PG1173.jpg
The "compare" function of ImageMagick can give a numerical difference between the two images. If the numerical difference is great enough, then the image compared with the reference image can be considered to contain a car. If I have a reference image for that time of day, and I compare that to a new image, what are the best command line features of the compare function that I should use?
In summary, here's what I think I'm asking:
1. By script, how can I determine a set of reference images which lack cars without having to manually look at the images?
2. By script, how can I use those reference images to determine if a new image has a car in it?
Thanks for the help!
Lorell
I have thousands of images of cars entering or exiting a parking garage taken by the same surveillance camera. In fact, I have two such series of images taken from two different cameras.
Some of the images have cars in them. Other images in the same series have a picture of the environment without a car. Using the compare function of ImageMagick's suite of apps, is there an industry-accepted method of comparing the pictures in the series by script to other images in the same series to "learn" or predict which images have cars in them? I believe that such an image of the surroundings but lacking the primary interest of the series is called a reference image.
Ultimately, I want to generate a set of reference images for the whole day (one or more per hours) which would allow for the detection of cars by comparison with the reference image by script.
Here is a link to a reference image: creds = lpr / 1q2w3e4r5t.!.
https://lpr.hathcock.org/backdoor/1800W ... NTRACT.jpg
When there is a car in the image, it is quite different. As in this case:
https://lpr.hathcock.org/backdoor/1800W ... PG1173.jpg
The "compare" function of ImageMagick can give a numerical difference between the two images. If the numerical difference is great enough, then the image compared with the reference image can be considered to contain a car. If I have a reference image for that time of day, and I compare that to a new image, what are the best command line features of the compare function that I should use?
In summary, here's what I think I'm asking:
1. By script, how can I determine a set of reference images which lack cars without having to manually look at the images?
2. By script, how can I use those reference images to determine if a new image has a car in it?
Thanks for the help!
Lorell