OCR Software to read only specific part of images
Budget: $250 – $750 USD
Need an OCR software to extract numbers from an Image. The images we have are called ‘Aerial Photographs’ , these are images taken from Aerial Cameras by means of flying all the way from 1950's to early 2000s. (Pre-Satellite era). The Digital Images are 24cm x 24cm, 15.5 microns. There is no text whatsoever on the complete image as the images are of world locations. Each Frame ( Image = frame from henceforth) has a frame number printed on one of the corners of the image. Top right, Bottom right, bottom left, Top Left are the most common in 99% cases. The frame numbers printed are 3-5 digits in most of the cases starting with a leading 0 (Ex: 01758 , 00025 etc.).
The Frame number pattern/font and at times color may not be the same for all, different frames have different type of font etc. The OCR program needs to be trained to do 3 things
1- Identify the location of the frame number and isolate that area to read the Frame number
2- Identify the Frame number with a very high accuracy, margin of error needs to be very low. Since identification of these numbers is in the foundation of huge huge process that lie ahead. If in case the program is not able to generate/read the correct number, the user should be notified about it (the interface and input format is discussed later)
3- Each image as its original file name such as 02001, 02002 02003 02004 and so on. The program needs to read the frame numbers and assign the Frame numbers next to the file numbers in a Database. We need this to rename all the files with the Frame numbers, this is not the scope of this software however writing this to inform what the plan is.
Technicalities
1- No specific interface is required, we use command line programming to run our work, so nothing fancy or nothing flashy needed since this is just going to be run in the back end.
2- The input is TIFF files. The program should be run in a manner that it will take the user’s input location of the files.
3- The program needs to be designed in a way that it may integrate with our Python based programs. This is something to be discussed later.
Summary
1- Input TIFF images
2- OCR reads and recognises Frame numbers
3- Puts the Frame numbers in a database against the File name
Please have a look at the sample images attached to get an idea what the input looks like.
Specific details and other minor tweaks or requirements will be discussed in detail once shortlisted.
The Frame number pattern/font and at times color may not be the same for all, different frames have different type of font etc. The OCR program needs to be trained to do 3 things
1- Identify the location of the frame number and isolate that area to read the Frame number
2- Identify the Frame number with a very high accuracy, margin of error needs to be very low. Since identification of these numbers is in the foundation of huge huge process that lie ahead. If in case the program is not able to generate/read the correct number, the user should be notified about it (the interface and input format is discussed later)
3- Each image as its original file name such as 02001, 02002 02003 02004 and so on. The program needs to read the frame numbers and assign the Frame numbers next to the file numbers in a Database. We need this to rename all the files with the Frame numbers, this is not the scope of this software however writing this to inform what the plan is.
Technicalities
1- No specific interface is required, we use command line programming to run our work, so nothing fancy or nothing flashy needed since this is just going to be run in the back end.
2- The input is TIFF files. The program should be run in a manner that it will take the user’s input location of the files.
3- The program needs to be designed in a way that it may integrate with our Python based programs. This is something to be discussed later.
Summary
1- Input TIFF images
2- OCR reads and recognises Frame numbers
3- Puts the Frame numbers in a database against the File name
Please have a look at the sample images attached to get an idea what the input looks like.
Specific details and other minor tweaks or requirements will be discussed in detail once shortlisted.