The increasingly complex road traffic environment has brought severe challenges to the safety of road traffic. In this paper, taking the accurate detection of road targets as the starting point, and combined with the current advanced convolution neural network technology, it analyzes and discusses the feature learning method based on feature fusion, and on this basis, it further proposes the target detection method based on feature fusion, and then it verifies the effectiveness of the proposed method through the simulation, thus, it provides a certain reference for road target detection in complex road traffic environment.