We present updated empirical radio surface-brightness-to-diameter ($\Sigma-D$) relation for supernova remnants (SNRs) in our Galaxy. Our original calibration sample of Galactic SNRs with independently determined distances (Pavlovic et al. 2013, hereafter Paper I) is reconsidered and updated with data which became available in the past two years. The orthogonal fitting procedure and probability-density-function-based (PDF) method are applied to the calibration sample in the $\log \Sigma - \log D$ plane. Orthogonal regression keeps $\Sigma-D$ and $D-\Sigma$ relations invariant. Our previous Monte Carlo simulations verified that the slopes of the empirical $\Sigma-D$ relation should be determined by using orthogonal regression. Updated calibration sample contains 65 shell SNRs. 6 new Galactic SNRs are added to the sample from Paper I, one is omitted and distances are changed for 10 SNRs. The slope derived here is slightly steeper ($\beta \approx 5.2$) than $\Sigma-D$ slope in Paper I ($\beta \approx 4.8$). The PDF method relies on data points density maps which can provide more reliable calibrations that preserve more information contained in the calibration sample. We estimate distances to five new faint Galactic SNRs discovered for the first time by Canadian Galactic Plane Survey and obtained distances of 2.3, 4.0, 1.3, 2.9 and 4.7 kiloparsecs for G108.5+11.0, G128.5+2.6, G149.5+3.2, G150.8+3.8 and G160.1$-$1.1, respectively. The updated empirical relation is used to estimate distances of 160 shell Galactic SNRs and new results change their distance scales up to 15 per cent, compared to results from Paper I. The PDF calculation can provide even few times higher or lower values in comparison with orthogonal fit, as it uses totally different approach. However, in average, this difference is 32, 24 and 18 per cent for mode, median and mean distances.