Recent advancements in spatial transcriptomics (ST) technologies have enabled the generation of gene expression profiles alongside histopathological images while preserving spatial context. Effectively integrating gene expression data with spatial information is crucial for accurately identifying spatial domains, a task that is essential for downstream analyses. Several methods have been developed to combine histopathological images with spatial transcriptomics data. However, these approaches often either use images solely to infer spatial relationships between spots or learn embeddings for gene expression and images separately, without fully integrating the information. To address these limitations, we propose a novel deep learning method, named STCON, for spatial domain identification. Specifically, we use Graph Convolutional Neural Network (GCNs) to extract embeddings from two patterns of gene expression profiles and histopathological images. We align gene expression and histopathological information in a low-dimensional space through contrastive learning, and then further optimize the alignment of image information with gene expression through cross-modal prediction. We evaluated STCON on four real spatial transcriptomics datasets. Experimental results demonstrate that STCON achieves competitive performance in spatial domain identification compared with four state-of-the-art methods. Moreover, STCON effectively detects spatially variable genes (SVGs) with enriched expression patterns in the identified domains. Overall, STCON represents a powerful and efficient computational framework for spatial domain identification in spatial transcriptomics data.