The advancement of logistics system intelligence has exposed limitations in traditional crane systems, particularly in cargo recognition and handling efficiency. To address these challenges, this paper presents the design and implementation of an intelligent logistics crane system based on the OpenMV vision module and the STM32F427 microcontroller, with a focus on integrated optimization for digital recognition and autonomous grasping tasks. The system adopts a modular architecture consisting of two primary components: the perception layer and the execution layer. The perception layer integrates the FOMO lightweight object detection algorithm with the MobileNetV2 0.35 network, deployed on the OpenMV platform to enable edge-side recognition and localization of digit-labeled targets. Additionally, TCRT5000 infrared sensors are employed to facilitate track following and docking control. The execution layer, coordinated by the STM32F427 main controller, governs the motion system driven by M2006 motors as well as the end-effector subsystem composed of stepper motors and servo actuators, thereby accomplishing precise object grasping and transportation. Communication between all subsystems is achieved through serial interfaces, constructing a closed-loop architecture encompassing recognition, decision-making, and execution. During the model deployment and experimental validation phases, model training and optimization were carried out using the Edge Impulse platform, ultimately achieving a recognition accuracy of 93.75% and an F1 score of 0.96. The system demonstrates strong real-time performance, robustness, and scalability. Experimental results confirm that the proposed system exhibits excellent operational stability and practical applicability in dynamic, multi-target environments, thereby offering an effective solution for intelligent embedded logistics systems.