To address the problem that the basic convolutional neural network is susceptible to background interference and weak expression of important features in recognizing bollworm in apple orchards, an identification method of bollworm in apple orchard based on MC-Mask R-CNN (Mish and CBAM - Mask R-CNN) is proposed. Firstly, on the basis of Haar's traditional neural network, the apple orchard images collected by multiple sites are initial segmented iteratively. The bollworm individual image samples are extracted, and the sample is expanded in multiple ways to obtain the expanded sample data set for deep learning. Secondly, an MC-Mask R-CNN feature extraction network is built and the Mish function is selected as the activation function to avoid vanishing gradient and gradient explosion during back propagation. And meanwhile, an attention mechanism module CBAM combining channel attention and spatial attention is introduced to improve the model's ability to express the characteristics of apple orchard bollworm, which is conducive to extracting deep feature information. Lastly, the Mask R-CNN is used as the control model and the recognition average precision is used as the evaluation index. The ablation test results show: integrating both the Mish activation function and the attention mechanism module into the feature extraction network can improve the recognition average precision of the model; the recognition average precision of the proposed MC-Mask R-CNN model reaches 97.69%. Contrasted with the Mask R-CNN, the recognition average precision is 5.08% higher. The results indicate that MC-Mask R-CNN can accurately and effectively identify apple orchard bollworm and provide technical support for the green protection and control of apple orchard diseases and insect pests.