According to my NoteBook, I have randomly compared 10 BScan images and found that they have the same value in the upper right corner, why is that?
\nNoteBook
According to my NoteBook, I have randomly compared 10 BScan images and found that they have the same value in the upper right corner, why is that?
\nNoteBook
Yeah you are right, the upper right part of the data are the same throughout the dataset. This has to do with the nature of the data. The Bscan (2D image) corresponds to the multi-offset received GPR signal. Essentially we have one transmitter at point x=0 and z=0, and many receivers at z=0 and at different x positions. Every column of the 2D image corresponds to the received radar signal at a specific receivers, and the x-position of the receiver is proportional to the index of the column. This means that the signal will arrive earlier at the first columns (receivers closer to the transmitter) and later at the last columns where the receivers are further away from the transmitter. Consequently there is going to be an area in each column where the signal hasn't arrived yet, i.e. the signal will be zero. Therefore there is a diagonal boundary in the Bscans, above which the signal will always be zero.
\nHope this was useful, let me know if any further clarifications are needed.
", "rawMarkdown": "Yeah you are right, the upper right part of the data are the same throughout the dataset. This has to do with the nature of the data. The Bscan (2D image) corresponds to the multi-offset received GPR signal. Essentially we have one transmitter at point x=0 and z=0, and many receivers at z=0 and at different x positions. Every column of the 2D image corresponds to the received radar signal at a specific receivers, and the x-position of the receiver is proportional to the index of the column. This means that the signal will arrive earlier at the first columns (receivers closer to the transmitter) and later at the last columns where the receivers are further away from the transmitter. Consequently there is going to be an area in each column where the signal hasn't arrived yet, i.e. the signal will be zero. Therefore there is a diagonal boundary in the Bscans, above which the signal will always be zero.\n\nHope this was useful, let me know if any further clarifications are needed.", "votes": null }, { "id": "3179063", "postDate": "04/15/2025 01:52:49", "content": "@iragia Thanks, that's helpful.
", "rawMarkdown": "iragia Thanks, that's helpful.", "votes": null }, { "id": "3179080", "postDate": "04/15/2025 02:34:14", "content": "@iragia I want to make sure that this pattern appears on all BScan images (not only in the training set but also in the test set)?
", "rawMarkdown": "iragia I want to make sure that this pattern appears on all BScan images (not only in the training set but also in the test set)?", "votes": null }, { "id": "3180480", "postDate": "04/16/2025 17:25:19", "content": "Yes this is correct, this pattern appears to all data, both training and testing data.
", "rawMarkdown": "Yes this is correct, this pattern appears to all data, both training and testing data.", "votes": null } ], "comments": [ { "id": 3178443, "author_name": "iragia", "author_url": "", "post_date": "04/14/2025 07:37:22", "content": "Yeah you are right, the upper right part of the data are the same throughout the dataset. This has to do with the nature of the data. The Bscan (2D image) corresponds to the multi-offset received GPR signal. Essentially we have one transmitter at point x=0 and z=0, and many receivers at z=0 and at different x positions. Every column of the 2D image corresponds to the received radar signal at a specific receivers, and the x-position of the receiver is proportional to the index of the column. This means that the signal will arrive earlier at the first columns (receivers closer to the transmitter) and later at the last columns where the receivers are further away from the transmitter. Consequently there is going to be an area in each column where the signal hasn't arrived yet, i.e. the signal will be zero. Therefore there is a diagonal boundary in the Bscans, above which the signal will always be zero.
\nHope this was useful, let me know if any further clarifications are needed.
", "votes": null, "replies": [ { "id": 3179063, "author_name": "ucas0v0zhuoqunli", "author_url": "", "post_date": "04/15/2025 01:52:49", "content": "@iragia Thanks, that's helpful.
", "votes": null, "replies": [] }, { "id": 3179080, "author_name": "ucas0v0zhuoqunli", "author_url": "", "post_date": "04/15/2025 02:34:14", "content": "@iragia I want to make sure that this pattern appears on all BScan images (not only in the training set but also in the test set)?
", "votes": null, "replies": [ { "id": 3180480, "author_name": "iragia", "author_url": "", "post_date": "04/16/2025 17:25:19", "content": "Yes this is correct, this pattern appears to all data, both training and testing data.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3178304": "According to my NoteBook, I have randomly compared 10 BScan images and found that they have the same value in the upper right corner, why is that? \n[NoteBook](https://www.kaggle.com/code/ucas0v0zhuoqunli/bscan-pattern)", "3178443": "Yeah you are right, the upper right part of the data are the same throughout the dataset. This has to do with the nature of the data. The Bscan (2D image) corresponds to the multi-offset received GPR signal. Essentially we have one transmitter at point x=0 and z=0, and many receivers at z=0 and at different x positions. Every column of the 2D image corresponds to the received radar signal at a specific receivers, and the x-position of the receiver is proportional to the index of the column. This means that the signal will arrive earlier at the first columns (receivers closer to the transmitter) and later at the last columns where the receivers are further away from the transmitter. Consequently there is going to be an area in each column where the signal hasn't arrived yet, i.e. the signal will be zero. Therefore there is a diagonal boundary in the Bscans, above which the signal will always be zero.\n\nHope this was useful, let me know if any further clarifications are needed.", "3179063": "iragia Thanks, that's helpful.", "3179080": "iragia I want to make sure that this pattern appears on all BScan images (not only in the training set but also in the test set)?", "3180480": "Yes this is correct, this pattern appears to all data, both training and testing data." }, "source": "meta" }