Research On Vehicle Detection And Ranging Model Based On Attention Mechanism

With the continuous development of artificial intelligence technology,smart car technology in the driverless scene has become an important direction for a new round of technological changes.As an important part of the driverless car,vehicle detection and ranging systems are of great significance for improving the reliability and safety of smart vehicles in the driverless environment.Therefore,how to achieve efficient and real-time vehicle detection and ranging has become one of the hot research topics.Traditional deep learning-based vehicle detection and vehicle distance estimation algorithms alleviate the above problems to a certain extent.However,there are still problems such as the loss of high-resolution features of the target object,insufficient feature fusion,difficulty in detecting small target objects,and insufficient accuracy of the traditional vehicle distance estimation method.In view of the above problems,this paper mainly makes the following work:First,this paper proposes an efficient multi-scale parallel vehicle detection model based on attention mechanism.The algorithm uses a multi-channel parallel multi-channel neural network structure.On this basis,a weighted feature fusion method is used.The weighted fusion is carried out according to the contribution of the multi-channel features to the network,and the jump connection operation is incorporated to ensure the loss of Lossless propagation..The simulation results show that,compared with mainstream vehicle detection models,the model proposed in this paper effectively solves the problems of high-resolution feature loss and insufficient feature fusion in the detection process,and achieves better detection results.Secondly,this paper proposes a vehicle detection model based on multi-scale feature fusion.Aiming at the defect of the single fusion dimension in the traditional multi-scale feature fusion module,the algorithm fully integrates the multi-scale features through the two dimensions of time and space.And on this basis,the in-degree node is deleted to complete the lightweight tailoring of the model.The simulation results show that compared with the mainstream multi-scale feature fusion module,the algorithm effectively solves the problem of difficult detection of small targets and achieves better detection results.Finally,this paper proposes a vehicle distance estimation model based on deep learning.Due to the poor real-time performance and insufficient accuracy of traditional vehicle distance estimation algorithms,and cannot achieve end-to-end distance estimation tasks,Therefore,this article uses deep learning methods to estimate and restore the depth of the image,learn the distance features,and realize the end-to-end distance estimation task.The simulation results show that the vehicle distance estimation model solves the problem of insufficient accuracy in the traditional monocular vision-based vehicle distance estimation method and improves the accuracy of the vehicle distance estimation.