Intelligent data compression: Incredible acquisition rates without information loss
Automated NDE of large parts produce a huge volume of information. Traditionally, gating has been used to reduce the information to just peak amplitude and position on every A-scan.
Although C- and D-scans can be produced from this data set, no further analysis is possible. B-scan imaging is preferred, but this would require the recording of all A-scan data.
Decimation is frequently used to decrease the data volume. However, if A-scans are decimated more than indicated by the Nyquist criterion, flaw indications can be lost, so that the data reduction gets rather limited.
All DASEL ultrasound systems include a hardware implemented data-compression algorithm, which is simple and quite effective. Data reduction rates are programmable and can be very high (above 100:1), without loss of peak amplitude and relative position in the resulting A-scan.
To this purpose, given a compression factor F (i.e., obtaining a data reduction of F:1), every A-scan of length L samples is decomposed in a set of G contiguous sequences {S1, S2,..., SG} every one containing 2F samples. For each segment, minimum an maximum values are obtained and stored, achieving a compression of 2F:2 (equal to F:1). The important feature is that the maximum and minimum amplitude within every sequence and their relative positions are kept.
Figure 1 shows an example, where a decimation by a factor F=16 of an A-scan of length L= 4000 samples, produces loss of data (those marked 1, 2 and 3 in a), while using the described algorithm (in b), all the indications are kept with their exact amplitude and relative position. In both cases, the resulting A-scan has 250 samples.
Fig. 1 – A-Scan compression 16:1 by (a) decimation and (b) by the NPLDR algorithm.
Read more at:
- A Full Featured Ultrasound NDE System in a Standard FPGA