A polarization-sensitive neuromorphic vision system (PNVS) that synchronously possesses the capacities of polarized light perception and neuromorphic computing enables the detection of potential information or hidden features. While developing a polarization-sensitive optoelectronic memristor presents an intriguing avenue for building the foundational hardware of PNVS, it has proven challenging. In this work, a polarization-sensitive optoelectronic memristor based on the in-plane anisotropic two-dimensional ReSe is proposed. Thanks to the meticulous device structure design, the angle-dependent synaptic plasticity of the device is demonstrated. Further, the image preprocessing and recognition functions are implemented under the circumstances of 0 and 90° polarized light, and the learning accuracy is improved from 81.1 to 86.8 and 90.4%, respectively. To reveal the mining capacity of hidden information, the surface flaw detection application is finally demonstrated through enhancement in the degree of linear polarization (DoLP). This study provides an approach to developing polarization-sensitive neuromorphic devices for future polarization vision systems used in intelligent vehicles and robot vision.