Aiming at the problem of low insect identification accuracy and difficult detection of small target insects against the complex background of cotton fields in Xinjiang, a lightweight insect detection model based on YOLOv5s (LID-YOLO) is proposed. First, the GhostNet network is used to replace the original cross-stage partial CSPDarknet53 network in the backbone, and the Slim-Neck module is used to improve the neck network, to achieve a lightweight model. Second, the fusion module BottleNet Transformer is introduced to reduce the number of model parameters and enhance the capability of network feature extraction to better detect small targets. Finally, the normalization-based attention module (NAM) is added to extract detail features by applying a sparse weight penalty to suppress non-significant weights and improve model accuracy. The experimental results show that compared with YOLOv5s, the LID-YOLO model reduces the number of parameters, calculations, and model weight by 30.9%, 45.6%, and 29.7% respectively. The accuracy rate of LID-YOLO model reached 97.4%, and detection speed was 55.25 FPS, which is 1% point and 2.62 FPS higher than that of the original YOLOv5s model. The LID-YOLO model not only ensures lightweight, but also improves detection accuracy to better meet the requirements of crop insect detection.