Plant phenotypic information plays an important role in the process of variety breeding. Maize intelligent breeding unmanned vehicle is an important vehicle for automatic detection of maize ear during the breeding. In order to achieve accurate and fast detection of maize ear in field environment, this study presents an object detection method for maize ear based on improved YOLOv8 model. Firstly, the maize ear dataset is constructed for the training of the maize ear detection model. Secondly, the CBAM attention mechanism is introduced and added to the backbone and neck networks to improve the feature extraction capability of the model. Finally, the original CIoU loss function is replaced by the Wise-IoU loss function, which improves the accuracy of the model detection and accelerate the convergence speed of the model. The experimental results show that the precision, recall and of the improved YOLOv8 model are 81.0%, 76.0% and 82.9%, respectively. In comparison to the original model, the improved model achieved a precision increase of 1.9%, a recall increase of 2.7%, and an increase of 3.1%. The improved model performs accurately and reliably in the maize ear detection task, provides a methodological reference for acquiring phenotypic information, and has potential for maize intelligent breeding unmanned vehicle applications.