Robust Long-range Obstacle Detection in Farmland via LiDAR-Camera Fusion and Cascaded Ground Filtering

This dataset is provided in the format of ROS bag files (.bag), containing multi-sensor perception data collected in real-world farmland scenarios. It is structured to support the core experiments presented in our study. Dataset Contents: Ground Filtering Scenarios: Data capturing uneven farmland terrain and unpaved roads to evaluate cascaded ground filtering performance. Data Fusion Qualitative Experiments: Sequences featuring single pedestrians, single tractors, and mixed human-machine scenarios to validate multimodal feature fusion. Detection Recall Experiments: Sparse point cloud data at mid-to-long distances designed for evaluating detection recall capabilities across varying ranges. Object Tracking Experiments: Data sequences covering continuous long-range target movement under varying conditions to verify multi-object tracking stability. Usage Instructions: Use the following command to play back the data streams: rosbag play <filename>.bag Use the following command to inspect detailed topic metadata, message types, and duration: rosbag info <filename>.bag