Acousto-optic tunable filter (AOTF) is a new type of light splitter with fast tuning, stable structure and portability. In this paper, a hyperspectral microscopic imaging system is constructed by combining non-collinear AOTF with optical inverted microscopy. The feasibility of data augmentation based on hyperspectral images for object detection of skin squamous cell carcinoma is studied. The hyperspectral images collected from unstained sections of skin squamous cell carcinoma are processed into dataset. At the same time, the mature open source object detection model is selected and trained for 20,000 times. Using the trained model to detect the lesion area of other unstained sections, it is found that the model trained by hyperspectral image dataset has a good ability to distinguish the non-lesion area, and there is no false detection. And the model has a relatively accurate detection ability for large lesion area, but the results of the model for small lesion area are not ideal. After analysis, it is considered that the number of samples can be increased firstly, especially in small lesions, and the same to the hyperspectral images. In addition, the model for lesion detection can be further optimized. By increasing the complexity of the model, the model can learn more details and information in the image during the training process. The preliminary results of the experiment prove that hyperspectral imaging is feasible for data augmentation of lesion object detection dataset. This paper provides a new method for the object detection data augmentation of skin squamous cell carcinoma.