Binocular Vision-Based Target Detection Algorithm

In the field of target detection, algorithms are challenged with multi-objective optimization problems in identifying detection targets, and it is also crucial to improve the recognition of small and insignificant targets. In this study, the Feffol network model is proposed to achieve excellent performance in complex detection tasks. In the feature extraction stage, the Feffol network model carefully chooses Efficient-v2 as the backbone network structure, which can provide a solid foundation for subsequent detection by taking advantage of its efficient feature extraction capability. In addition, an Ebifpn feature pyramid module with an spatial pyramid pooling-cross-stage partial connections(SppCSP) structure was innovatively introduced. This design not only effectively expands the feature sensing field, but also greatly enhances the fusion and expression of feature information of different sizes, so that the model can capture the target features more comprehensively. In addition, to solve the key problems in the detection process, the Feffol model adopts the Focal Loss and CIoU Loss functions. The former can effectively balance the positive and negative samples to avoid over-learning of easy-to-classify samples during the training process; the latter successfully solves the problem of the failure of the detection method when there is no intersection between the prediction frame and the real detection frame, which significantly improves the robustness of the model. The experimental results show that the Feffol network model has excellent performance, with a detection accuracy as high as 90.09% and a detection speed of 9.75it/s. Compared with other mainstream networks, the detection accuracy of Feffol network has been significantly improved, which fully proves its advancement and effectiveness in the task of target detection, and provides new ideas and methods for the further development of this field.