Research on Multi-Feature Fusion of Improved Concat Module Based on Yolov11 in Wheat Spike Recognition

In order to improve the accuracy of wheat ear recognition, this study proposed a lightweight hybrid network Spike-YoloNet based on improved YOLOv11n. Aiming at the problem of high computational complexity of traditional CBAM, this study designed LightweightCBAM module, which can improve the performance and reduce the overhead. In addition, by constructing the AttentionConcat module and embedding LightweightCBAM in the Concat module of the original feature fusion layer of YOLOv11, the efficiency of feature fusion and the ability of identifying key information are effectively improved. The results show that the improved model increases F1 value by 1 percentage point and mAP@0.5 by 2 percentage points on the premise of keeping the parameters equal to the original YOLOv11n. Experiments show that Spike-YoloNet provides an effective solution for the development of wheat ear detection equipment with limited computing power.