Design And Realization Of Key Technologies Of Intelligent Warehousing In The Commodity Market

In recent years,with the increase of e-commerce platforms,the Yiwu small commodity market has also developed rapidly.Unlike express parcel storage warehouses,small commodity warehouses have problems such as fast in and out of goods,inconsistent shapes and sizes of goods,and limited capacity of warehouses built by merchants.The deployment method of large warehouses is not suitable for this special storage system.For example,in the case of large warehouses,the tracked roadway stacker is expensive to deploy and is not suitable for deployment in small merchant warehouses.In this paper,under the research background of the small commodity intelligent storage system,taking into account the characteristics of the small commodity intelligent storage system,the key technology in the intelligent storage system under this situation is designed and implemented.This article is mainly to improve the performance of the intelligent storage system in the small commodity scenario and the selection of key algorithms to improve the efficiency of the storage system.The main work of this paper is as follows:(1)The realization of the item detection algorithm,considering that items in the small commodity market have the characteristics of fast in and out,it is necessary to choose an algorithm with fast recognition speed and low recognition load in the selection of the recognition algorithm.In order to ensure the accuracy of the AGV's recognition algorithm,a monocular camera is mounted on the AGV as the AGV picture acquisition module,through which the target object is photographed.After the image collection is completed,this article compares the two mainstream image detection algorithms,the feature point detection algorithm and the YOLOv3 algorithm based on deep learning,and chooses the selection of the recognition algorithm suitable for the smart warehouse in the small commodity scene.First,use the SURF feature point extraction algorithm to identify the feature points of the picture to be recognized,and then use the MSAC algorithm as the filtering algorithm to filter the feature points,and screen out the points with strong characteristics.Finally,the filtered feature points are compared with the previous Match the registered points in the database.Because of the high requirements for real-time and fast detection in the storage environment,the algorithm needs to traverse the entire database every time it matches.Therefore,the matching algorithm is not suitable for registered commodities.When there are too many warehousing environments,on this basis,we use the YOLOv3 algorithm based on the deep learning framework.Because the YOLOv3 algorithm has the advantages of high recognition accuracy,fast recognition speed and strong generalization ability,we choose this algorithm as an intelligent warehousing in small commodity scenarios.The algorithm model in the system's target detection subsystem.This article divides the items to be tested into ten categories in the training set.Each category selects more than 200 pictures,a total of 2121 pictures for training.When the item parameters are used as the model training,input After the target picture,the target detection can be accurately realized.(2)Research on indoor positioning technology.AGV needs to report its location to the control terminal in real time,so that the control terminal can issue tasks.This article first compares four common indoor positioning technologies,and finally chooses UWB positioning technology as the research method.The performance comparison of the Fang,Chan and Taylor algorithms based on the time difference of arrival method is studied,and the results are finally obtained.(3)In the data processing system,the terminal designs the optimal path to the target point after receiving the AGV's positioning feedback.In response to this problem,this paper improves an improved algorithm based on the A* algorithm,taking into account that the warehouse has a relatively high In the case of multiple obstacles,first introduce the obstacle rate into the heuristic function to ensure the speed of the algorithm,on the other hand,to ensure that the path calculated by the algorithm is globally optimal;in order to avoid the risk of collision,consider the volume of the automatic guided car to introduce an anti-collision strategy This can make the movement path smoother and prevent collisions.Finally,the algorithm is compared with Dijkstra algorithm,ant colony algorithm,and A* algorithm,and finally the result is obtained.The test results show that the improved A* algorithm performs well regardless of the running time of the algorithm,the number of nodes searched,the length of the planned path,and the number of turns to guide the car.