In the present study, a laboratory soil-cracking experiment was performed
on 56 soda-saline soil samples. Correlation analyses between
one GLCM texture feature (CON) and six main salt parameters (EC,
Na+ content, Cl− content, CO3
2− content, HCO3
− content and salinity)
were conducted. Correlation analyses between the CON value and the
main salt parameters of 48 field soil samples were also conducted. The
CON values of the crack images were significantly positively correlated
to the EC, Na+ content and salinity (R2 N0.87) and linearly related to
the Cl− content (R2 N0.73). An on-line measurement method of determining
soil salt parameters was then proposed based on calculations
of the CON values of binary images of naturally cracked soil. After that
the salt parameters of 48 field verification points were estimated
using the on-line measurement method. The R2 and RPD of the fitting
results (to a function of y = x) between the estimated values and measured
values indicates that the on-line measurement method used in
the present study has relatively high accuracy for predictions of certain
parameters, such as the EC, Na+ content, Cl− content and salinity.