The problem of source number detection based on a distributed polarization-sensitive array (DPSA) is considered in this paper. We propose a DPSA to improve the matching coefficient that can be used to measure the eigenvalues. We theoretically prove that DPSA has a lower matching coefficient than the traditional uniform circular array (UCA) when the incident sources are uncorrelated, and we then obtain the eigenvalues by using the matching coefficient. Moreover, we discuss the change in eigenvalues under the assumption that the power of the signals is unequal. Numerical simulations show that the source detection probability of the DPSA is enhanced compared with the traditional UCA. Meanwhile, increasing the power of the stronger received signal improves the detection probability when the source power is unequal.