To address the challenge of automated detection of void of the steel-concrete structures interfaces, this paper designs a magnetic wall-climbing robot incorporating image and acoustic feature recognition. First, the chassis, the magnetic suction device, the power system, the drive system, the mapping and sound modules of the wall-climbing robot are introduced. Secondly, the hardware control system is highlighted, and the feasibility of Neodymium-iron-boron permanent magnet as an adsorption force is determined through force analysis and experimental testing. Finally, the composition and functions of the image and acoustic feature recognition software part are detailed. Among them, the image capturing part adopts the image transmission solution based on the Orange Pie platform. The acoustic pattern recognition module consists of front-end, middle-end and back-end architecture: the front-end carries percussion and recording devices for excitation and collection of acoustic patterns; the middle-end of the acoustic pattern recognition WeChat mini program is developed to achieve acoustic pattern noise removal and effective feature extraction; the back-end, through the cooperation of Tencent Cloud and the WeChat mini program, recognizes acoustic pattern data and returns the results to the WeChat mini program. The magnetic wall-climbing robot incorporating image and acoustic feature recognition designed in this paper can achieve the collaborative acquisition and analysis of image and acoustic patterns, providing an effective solution for the automated inspection of steel-concrete and other structural interfaces for debonding.