Ship Recognition For Inland Port Transportation Management

Chongqing is located at the intersection of the “Yangtze River Economic Belt” and the “Silk Road Economic Belt”.It is the largest shipping logistics hub and passenger transportation centre in the middle and upper reaches of the Yangtze River,and shipping is very busy.The traditional information management method can no longer meet the needs of port ship traffic statistics,berth guidance and supervision,illegal transportation along the river,illegal(or irregularities)sand mining and the inspection and protection of important water source rivers.This paper is based on actual projects for port ship berthing guidance and important water source inspection and protection needs.Make full use of the difference between frame and frame for the stability of slow moving target detection,and The advantages of high efficiency and easy realization of background difference operation.The inter-frame difference operation and the background difference operation are respectively performed on the adjacent 6-frame images.Finally,take two methods to get the union of the pictures.Therefore,a ship adaptive detection method with 6 frames of frame difference and background difference is proposed.The experimental results show that the method has high adaptive detection accuracy for ships with slow inland navigation.Make full use of the illumination invariance of the color name feature,the monotonic gray invariance of the local binary feature(LBP),and the good local texture expression ability combined with the geometric and optical advantages of the HOG feature.Multi-feature fusion kernel correlation filtering ship tracking method based on weighted feature fusion.The method avoids the disadvantage that the single feature kernel correlation filter tracking algorithm tracks accurately and lowly.An effective visual calculation method for automatically determining whether there is a ship entering the parking field and its parking time.Automatic identification of ship types for port traffic statistics,berthing guidance,and channel inspection.Make full use of the advantages of fast learning speed of extreme learning machines.Improve the classification performance of CNN,and use the extreme learning machine instead of CNN's softmax function.An identification method combining CNN and extreme learning machine is proposed.Combined with the ship picture data set provided by Chongqing Port Transport Bureau.The experimental results show that the error rate of ship classification and recognition is reduced to 9.49%,and the accuracy and efficiency of classification recognition are improved obviously.