Oil-water biphasic medium widely existed in mining and transportation process of petroleum industry, where visualized measurement processes vital importance in production increase, state monitoring and fault prognosis. Ultrasonic Tomography (UT) is a non-invasive and non-radioactive sensing modality in industrial Process Tomography, which is demonstrated as a competitive monitoring technique with great potential. However, facing the refraction effect of Ultrasound wave at oil-water interfaces, the accuracy of reconstruction algorithm is with large error and serious artifacts. To address the issue, a modified ray-tracing image reconstruction algorithm is proposed. The amplitude attenuation during ultrasound propagation in nonhomogeneous medium is quantitatively modeled to form absolute UT model. With ray tracing strategy applied and improved, the nonlinear refraction effect is characterized in the form of iterative update of sensitivity matrix, followed by which the sound speed mapping transfers reconstructed attenuation images to sound speed distribution for ray tracing computation. Numerical tests indicate that the proposed method significantly improves reconstruction accuracy through projection-based, regularization-based and algebraic-based inverse algorithms, which demonstrate its benefits in conducting oil-water biphasic medium imaging.