Tomosynthesis Height Estimation for PCB Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional tomographic methods face limitations in accurately determining the depth of imaged structures, particularly in non-destructive testing of objects like multilayer printed circuit boards or welding seams, due to background intensity reducing contrast resolution and requiring complex mechanical scanning systems.
Innovation Solution
A method and system utilizing tomosynthesis-like techniques to estimate the height profile of objects by acquiring and processing raw images with spatial shifts, calculating volatility parameters, and transforming image shifts into height values, enabling the generation of a height profile image and detection of defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simple irradiation techniques are used, then the measurement process is simple, but there is no possibility to get information about the depth of the imaged structures
Solution Approach 1:
The patent transitions from 2D projection imaging to 3D depth information by introducing the time dimension through temporal sequences of images. Multiple images captured at different time points are processed to reconstruct depth profiles, effectively adding a dimensional aspect to the measurement capability without requiring complex multi-angle mechanical scanning systems
2Measurement precision
If classical laminography is used to image object slices, then depth position can be determined, but background intensity reduces contrast resolution
Solution Approach 1:
The patent extracts only the relevant depth information from the temporal image sequences while discarding the problematic background intensity components. By processing images through temporal correlation and depth reconstruction algorithms, the method isolates the focal slice information from out-of-focus background structures, thereby maintaining contrast resolution while achieving depth positioning
3Measurement precision
If rotational laminography is used to obtain projections from bigger angular region, then better imaging results are achieved, but a more complicated mechanical scanning system is required
Solution Approach 1:
The patent replaces complex mechanical rotational scanning with a simpler temporal imaging approach. Instead of physically rotating the x-ray source and detector through large angular regions, the method captures a sequence of images during linear translation and uses temporal processing to achieve depth resolution, thereby substituting mechanical complexity with computational processing
4Measurement precision
If computed tomography is used for nondestructive imaging of object slices, then imaging accuracy is improved, but objects must be irradiated from the full angular region which is not always possible
Solution Approach 1:
The patent achieves slice imaging capability without full angular coverage by introducing temporal dimension. A sequence of images captured during linear translation provides sufficient information for depth reconstruction, eliminating the requirement for objects to be accessible from all angles while maintaining imaging accuracy for flat components
Data Source
AI summary
Systems and methods to estimate the height profile of an object using tomosynthesis-like techniques. A plurality of raw images of an object to be characterized are acquired, where the plurality of raw images are representative of a plurality of spatial shifts of an imaging device relative to the object to be characterized. The raw images are processed to generate composite images, where each composite image corresponds to a unique image shift between spatially adjacent raw images. A volatility parameter value is calculated within a neighborhood of a same image pixel location for each composite image. The composite image having the largest volatility parameter value for the image pixel location is determined. A unique image shift, corresponding to the composite image having the largest volatility parameter value, is transformed into a height value representative of a height dimension of the image pixel location.