Techniques to reconstruct data from acoustically constructed images using machine learning
Acoustic data, such as a full matrix capture (FMC) matrix, can be reconstructed by applying a previously trained decoder machine learning model to one or more encoded acoustic images, such as the TFM image(s), to generate reconstructed acoustic data. A processor can use the reconstructed acoustic data, such as an FMC matrix, to recreate new encoded acoustic images, such as TFM image(s), using different generation parameters (e.g., acoustic velocity, part thickness, acoustic mode, etc.).
1. A computerized method of image processing using processing circuitry to apply a previously trained decoder machine learning model in a system for non-destructive testing (NDT) of a material, the method comprising:
obtaining a first encoded acoustic image of an object, wherein the first encoded acoustic image is generated using raw acoustic data and at least one generation parameter and at least one inspection parameter, wherein the at least one generation parameter includes at least one of wedge velocity, wedge angle, wedge first elements height, wedge radius, part velocity, part thickness, and selected acoustic mode;
applying a previously trained decoder machine learning model to the first encoded acoustic image to generate reconstructed raw acoustic data;
modifying the at least one generation parameter; and
applying the modified at least one generation parameter and the reconstructed raw acoustic data to an encoder to generate a second encoded acoustic image of the object.
2. The computerized method of claim 1 , comprising:
performing acoustic acquisition techniques to obtain the raw acoustic data.
3. The computerized method of claim 2 , wherein performing acoustic acquisition techniques to obtain the raw acoustic data includes:
performing full matrix capture techniques to obtain the raw acoustic data.
4. The computerized method of claim 2 , wherein performing acoustic acquisition techniques to obtain the raw acoustic data includes:
performing virtual source aperture techniques to obtain the raw acoustic data.
5. The computerized method of claim 1 , wherein the first encoded acoustic image is generated using a total focusing method (TFM) technique.
6. The computerized method of claim 1 , wherein the first encoded acoustic image is generated using a synthetic aperture focusing technique (SAFT).
7. The computerized method of claim 1 , comprising:
selecting the previously trained decoder machine learning model from a plurality of previously trained decoder machine learning models.
8. The computerized method of claim 7 , wherein the plurality of previously trained decoder machine learning models were trained using different generation parameters.
9. The computerized method of claim 1 , wherein both the first encoded acoustic image and the raw acoustic data include three-dimensional (3D) data.
10. A computerized method of image processing using image processing circuitry to apply a previously trained decoder machine learning model in a system for non-destructive testing (NDT) of a material, the method comprising:
obtaining a first encoded acoustic image of an object, wherein the first encoded acoustic image is generated using raw acoustic data and at least one generation parameter and at least one inspection parameter, wherein the at least one generation parameter includes at least one of wedge velocity, wedge angle, wedge first elements height, wedge radius, part velocity, part thickness, and selected acoustic mode;
modifying the at least one generation parameter; and
applying a previously trained decoder machine learning model to the first encoded acoustic image to generate a second encoded acoustic image using reconstructed raw acoustic data.
11. The computerized method of claim 10 , comprising:
performing acoustic acquisition techniques to obtain the first acoustic imaging data.
12. The computerized method of claim 11 , wherein performing acoustic acquisition techniques to obtain the raw acoustic data includes:
performing full matrix capture techniques to obtain the raw acoustic data.
13. The computerized method of claim 11 , wherein performing acoustic acquisition techniques to obtain the raw acoustic data includes:
performing virtual source aperture techniques to obtain the raw acoustic data.
14. The computerized method of claim 10 , wherein the first encoded acoustic image is generated using a total focusing method (TFM) technique.
15. The computerized method of claim 10 , wherein the first encoded acoustic image is generated using a synthetic aperture focusing technique (SAFT).
16. The computerized method of claim 10 , wherein both the first encoded acoustic image and the raw acoustic data include three-dimensional (3D) data.
17. A computerized method of image processing using image processing circuitry to apply a previously trained decoder machine learning model in a system for non-destructive testing (NDT) of a material, the method comprising:
obtaining a first encoded acoustic image of an object, wherein the first encoded acoustic image is generated using raw acoustic data and at least one generation parameter and at least one inspection parameter, wherein the at least one generation parameter includes at least one of wedge velocity, wedge angle, wedge first elements height, wedge radius, part velocity, part thickness, and selected acoustic mode;
downsampling the first raw acoustic data and storing the downsampled raw acoustic data;
applying a previously trained decoder machine learning model to the downsampled raw acoustic data and the first encoded acoustic image to generate reconstructed raw acoustic data;
modifying at least one generation parameter; and
applying the modified at least one generation parameter and the reconstructed raw acoustic data to an encoder to generate a second encoded acoustic image of the object.
18. The computerized method of claim 17 , wherein downsampling the raw acoustic data and storing the downsampled raw acoustic data includes:
removing amplitude information of the raw acoustic data and storing phase information of the raw acoustic data.