Research on Adaptive Weight and Frequency Domain Enhancement Fusion Method for Small Target Detection

The detection of small targets is very important in computer vision, particularly in applications like drone surveillance, medical image interpretation, and smart transportation. However, due to the extremely small proportion of small targets in the image, traditional detection models often face high recognition difficulty and high false detection rate when dealing with these tasks. With the rapid development of deep learning, the YOLO series of models have become the mainstream target detection method due to their efficient detection speed and good accuracy performance. This research enhances small target detection performance by introducing the adaptive feature weight allocation algorithm (AFW) and the frequency domain enhancement module (MEA) based on YOLOv10. The enhanced model demonstrates significant benefits across various metrics in experimental findings, particularly excelling in identifying small targets of varying sizes and intricate backgrounds, displaying superior resilience and adaptability. The results of the ablation experiment also indicate that the combination of AFW and MEA modules can significantly enhance the detection accuracy. This research offers a very adaptable and useful solution for identifying small targets and offers a useful guide for the future advancement of small target detection technology.