The Mask R-CNN (mask region-based convolutional neural network) object detection method is proposed based on the improved feature pyramid. The experimental results show that compared with the Mask R-CNN detection structure, the mean average precision (mAP) under different Intersection-over-Union (IoU) thresholds increases by 2.4% and 3.8% in the detection of object edge and bounding box, respectively. In particular, the detection accuracy of medium size objects is greatly improved by 7.7% and 8.5%, respectively, which indicates strong robustness.