Addressing the challenge of large-scale metal debris collection in desert environments, we have developed an intelligent robot capable of autonomous detection and retrieval of metal fragments. Utilizing a behavior tree-based control framework, we propose a multi-sensor fusion algorithm for metal debris detection. We have encapsulated fundamental functionalities such as localization and navigation, target recognition, and robotic arm control. The robot's various modules have been integrated, and real-world application and optimization have been conducted. Test results indicate that the robot exhibits a high success rate and efficiency in metal debris collection in open desert environments, along with notable stability and a low failure rate.