During microlens array（MLA） photolithography exposure process， the number of photolithography points is considerably large， thus， judgement of the photolithography quality by human eyes with a high-magnification microscope is time-consuming and labor-intensive， resulting in high process cost. To solve this problem， an easily detected circular pattern was designed and a Yolov5 model for target detection in deep learning was introduced， which can replace manual eye inspection to a certain extent and complete the rapid judgement of photolithography quality. Based on the proposed method， the optimal interval of the level of energy density during laser scanning and the profile dip angle of the photoresist were analyzed under different photoresist thicknesses. At the same level of energy density during laser scanning， the distortion of photolithography pattern was judged considering circularity. Further， the photoresist thickness， laser power， and processing platform moving speed were selected as independent variables in the MLA photolithography process to evaluate processing quality parameters processing quality parameters， such as photolithography qualification rate， photoresist profile inclination angle， and photolithography circularity， is of great significance for engineering.