Pre-terminated optical distribution box is a kind of significant optical transmission device, which is in charge of the connection and management of the physical distribution of optical fiber network. Due to the information recognition of the box mainly relies on manual check nowadays, it is urgent to develop intelligent way to reduce costs and improve efficiency. Nevertheless, due to the complex illumination, diverse port states and the occlusion between fibers, existing computer vision technologies have difficulties to recognize the port information (port states and barcodes of the optical-fibers), which specifically includes inaccurate recognition and clueless mapping between the identified states and corresponding physical positions. To tackle with such challenges, an automatic port information recognition method based on spatial matching and calibration for pre-terminated optical distribution box is proposed, including: (1) a port information recognition and calibration method based on spatial multi-scale matching and multi-angle compensation. The multi-scale template calibration is designed to adaptively map and assign port instances to their responding physical locations, and the multi-angle compensation is responsible for calibrating and recognizing of the missed ports. (2) an optical fiber barcode identification algorithm based on adaptive correction and dynamic minimum mapping. Specifically, the barcode is successively segmented according to its imaging posture and is assigned to the fiber port by spatial matching and calibration guided by dynamic minimum matching, which achieves the recognition of the barcode. Moreover, we construct a real dataset, and on which we carry out abundant experiments to demonstrate the superiority of the proposed scheme in both accuracy and efficiency in port information recognition. At the same time, the scheme is deployed and tested on mobile devices, realizing accurate and efficient automatic port recognition on pre-terminated optical distribution box.