To ensure the engineering quality and durability of reinforced concrete, vibrating is necessary during the pouring process. Currently, the vibration process during the construction of high-rise buildings is mainly done manually using handheld vibrating rods. However, manual vibration relies heavily on human experience and often results in issues such as over-vibration or insufficient vibration. Once the concrete solidifies, these problems can lead to irreversible consequences. This paper proposes a vibration robot that replaces manual vibrating with a strictly programmed process, improving construction efficiency and project quality. The vibration robot is based on the YOLOv5s algorithm and incorporates Deformable Convolutional Networks (DCN) and Coordinate Attention (CA) mechanisms into the backbone network to enhance target recognition performance. Through image recognition, the vibration robot obtains the coordinates of the vibration center point for controlling the vibration operation. Experimental results show that this method achieves a target recognition rate of 95.24% and a frame transfer rate of 16.94 frames per second. Therefore, using a vibration robot for concrete vibration during pouring holds significant practical value.