Aiming at the contradiction between real-time requirements and resource constraints of crop pest identification in UAV edge computing scenarios, this research proposes a systematic solution based on TRIZ theory. By using the contradiction matrix and object field model, the two core problems of the conflict between high-resolution image requirements and drone endurance, as well as the imbalance between model lightweighting and accuracy, have been solved: designing a dynamic sampling strategy for sub regions to optimize aerial energy consumption, constructing a lightweight dual stream network architecture to achieve edge deployment, and combining multispectral fusion and adaptive attention mechanism to enhance the ability to extract small target features. The 13 proposed solutions are comprehensively evaluated from two aspects of technical feasibility and economy, and finally a technically feasible and economically economical solution for pest identification under UAV edge computing is obtained, which points out the direction of technological breakthrough for subsequent practical research. The experiment shows that the average precision of the model on the self built dataset is 94.7%, 12.3% higher than that of the traditional method. The model can be integrated into the agricultural UAV platform, providing an efficient solution for edge computing scenarios.