Surgical guidance based on spectral imaging can effectively identify various tissue types by analyzing the spectrum differences of different tissues, which has a significant application value. However, the current sampling speed of spectral imaging and the presentation of spectral images severely limit its clinical applications. This paper proposes an augmented reality (AR) computational spectral imaging system for facilitating surgical guidance. High-quality spectral reconstruction is achieved in a single detection using RGB imaging devices and multispectral single-pixel detectors, which greatly improves the imaging speed and reconstruction efficiency. Principal component analysis and spectral angle mapping are adopted to extract the effective information of the spectral images, which are then used to highlight the characteristic tissue area. The image enhancement effect based on the fusion of the spectral image and actual surgical area is demonstrated in the head-mounted AR display device.