Multispectral image fusion and ant colony optimization for path planning of farm inspection robots

The autonomous navigation and environmental perception capabilities of Farm Inspection Robots (FIRs) are crucial for achieving precision agriculture. To improve their navigation performance and perception accuracy in complex and dynamic farmland environments, this study uses Poisson Image Editing (PIE) to achieve seamless fusion of visible and Thermal Infrared Images (TIIs). It combines an improved Conditional Generative Adversarial Network (CGAN) to enhance multi-scale feature extraction and edge preservation capabilities. In terms of path planning, an improved Ant Colony Optimization (ACO) algorithm is used to optimize the global path, and the Dynamic Window Approach (DWA) is combined to achieve local dynamic obstacle avoidance, aiming to balance global optimality and real-time response. The results show that in terms of image fusion, the edge strength of the improved CGAN reaches 96.02, the average gradient is 0.0865, and the information entropy is 6.81, all of which are superior to the comparison methods. In terms of path planning, the path length of the improved ACO in a 30 × 30 grid map is 47.85 m, with 10 turns, which is 6.2% shorter than the traditional ACO, and the convergence speed is increased by 34.5%. After combining with DWA, the robot can successfully avoid dynamic pedestrians and static obstacles and reach the target point without collision. Through the collaborative optimization of multi-spectral image fusion and path planning, this study effectively improves the autonomous navigation performance and operation reliability of FIRs in complex scenarios, providing key technical support for promoting precision agriculture, reducing labor costs, and achieving intelligent farm management.