Figure 5. 1-s spectrum at different positions
From figure 5, when the detector was located at the position with high dose rate, such as the position 10 cm away the source, there exist obvious signal at the channel range where the characteristic peak located. When it was located at the low dose rate positions, the signal was similar with the background.
Rapid detection method of radioactive source based on low-count spectra
From the above investigation result, only the spectrum containing high dose rate signal can be detected from 1-s spectrum. This undoubtedly reduces the detection sensitivity greatly. In the actual application, it will increase the risk that worker exposed to the radiation environment. From the perspective of radiation protection safety, we should ensure that nuclear workers receive as little radiation dose as possible. Therefore, we should develop a detector with high sensitivity and low false alarm rate. However, the two demands are a contradiction. We need to find the best compromise between them.
Since the application scenario of such detector is site walking patrol, according to the pedestrian speed, the abnormal signal should be determined within 5 seconds to avoid the missing alarm. From the above test result of the 100-s mean spectra, it is unable to discern the source contained spectra and background spectra according to only 5-s spectra obviously.
This paper proposed a rapid detection method of radioactive source based on low-count spectra. The key technology of the method was principal component analysis (PCA), which was able to extract the feature value of the spectra and help to discern the 1-s source contained spectra and 1-s background spectra easily. The method included the follow steps.
Step 1. 60~100 s background spectra were collected as the background spectra matrix.
Step 2. Constructed the feature extraction matrix with the new measured 1-s spectrum and background spectra matrix, then acquired the maximum feature coefficient with PCA.
Step 3. The maximum value in the range of (240,290) was set as the feature value of the spectrum.
Step 4. Determine whether source contained spectrum were detected based on the feature values of 5 consecutive seconds spectra.