| Vehicle autonomous navigation algorithms are currently a research hotspot in the field of intelligent vehicles,traditional vehicle autonomous navigation systems are highly dependent on GPS location signals.However,in modern military environment combat scenarios,it is often difficult to obtain accurate and reliable GPS signals due to signal interference and other factors;at the same time,due to signal interference,it is difficult for vehicles to obtain real-time information about the trafficability of the road ahead,which is difficult to meet navigation requirements in the complex and changeable environment.Therefore,it is of great strategic significance to realize the autonomous navigation of vehicles in complex environments without GPS.In response to the above problems,this paper proposes a new vehicle autonomous navigation algorithm to provide military intelligent vehicles with location,road trafficability and route planning information in a without GPS and complex environment.Firstly,in order to identify the global position information of the vehicle in a GPS-free environment,a new vehicle visual position recognition algorithm is proposed through the matching retrieval of aerial images and offline satellite images.The algorithm introduces the attention mechanism and fine-grained feature extraction module on the basis of the twin network structure,mines the discriminative features in satellite and aerial images,and trains the above model using the cross-entropy loss function of smooth labels.The simulation experiment results show that the average accuracy and recall rate of the location recognition algorithm reach 75.37% and71.64%,respectively,which has high reliability.Secondly,on the basis of obtaining vehicle location information,an elite ant colony optimization algorithm is used to perform global path planning and path re-planning under impassable conditions.By improving the pheromone update strategy and heuristic function of the ant colony algorithm,the ant colony algorithm search method is optimized,and the search speed of the optimal path is improved.Establish a model in MATLAB simulation software to simulate and analyze the above algorithm and the elite ant colony algorithm.The results show that the algorithm in this paper has a faster convergence speed and a strong optimization ability.After that,aiming at the identification of road accessibility in complex road environments,lidar is used to detect the positive obstacles,negative obstacles,road edges and other environmental information on the road ahead.The random sampling consensus algorithm is used to iteratively fit the ground point cloud segmentation to achieve the segmentation of the ground point cloud and the positive obstacle point cloud,and the negative obstacle detection is performed by calculating the amplitude interval of the laser points on the same scan line.At the same time,the multi-feature fusion method is used for each laser point on the laser line to extract the boundary points on both sides of the road,and through B-spline curve fitting,the road boundary detection is realized.The simulation results show that the algorithm can detect the road boundary and obstacles within the road boundary on the KITTI lidar data set,thereby effectively identifying road traffic.Finally,in order to further verify the effectiveness of the vehicle navigation algorithm,collect aerial images of the actual environment of a university campus,and perform visual position recognition based on the above network model;use the vehicle-mounted 16-line lidar to verify the road negative obstacle recognition algorithm;finally establish the actual environmental road Net map,based on the elite ant colony optimization algorithm for global path planning.The results show that the algorithm can provide safe and reliable autonomous navigation information for military intelligent vehicles without relying on GPS-assisted positioning,and has good engineering practical value. |
