Research On Calculation Method Of Regular Item Volume Based On Two-dimensional Image

As the Internet and electronic Commerce developing rapidly in recent years,logistics industry has transformed from traditional labor-intensive to technology-intensive smart logistics.The volume of regular objects is a crucial data in smart logistics.Compared with weight and other basic data,it is more difficult to be obtained.Whether it is a single regular object or multiple irregular objects stacked together,it's valuable to research how to obtain volume accurately and conveniently.This dissertation takes the most common carton in logistics as the research object,researching volume measurement from a single regular object and multiple stacked regular objects respectively.To meet the needs of volume measurement,the key-point for single regular object is improving the accuracy of 3D Reconstruction in condition of rarely raising the network parameters.The volume measurement of multiple stacked regular objects mainly focuses on how to deduct the excess part of the external cube of the stacked objects through several fixed cameras,and accurately calculate the volume of the stacked objects.To improve the measurement performance for mentioned problems,this dissertation focuses on the research work and results listed below:1.Current volume measurement algorithms for single regular object have draw-backs such as the limiting use scenario,the high cost,and the low accuracy,this dis-sertation innovatively proposes a singleview 3D reconstruction volume measurement algorithm,without the use of auxiliary hardware equipment,which reduces costs and facilitates couriers to obtain accurate volume data anytime,anywhere.In this dissertation,the latest instance segmentation algorithm is used to quickly and accurately segment the foreground in the image and remove the irrelevant background.In addition,for the problem of low-resolution based on single-view 3D reconstruction algorithm,this dissertation improves the output resolution of the reconstruction algorithm while reducing the amount of network parameters,and subsequently corrects the reconstructed voxel model through a noise removal algorithm,thereby get a more refined reconstruction model.The experimental results prove that compared with the previous algorithm,the improved 3D reconstruction algorithm has improved performance,and the volume measurement algorithm has the advantages of convenient use,no restrictions on the use scenarios,higher accuracy and other advantages.The final average accuracy of the algorithm reaches 4.02%,and meets the needs of actual engineering.2.Current volume measurement algorithm for regular object stacking in logistics,most volume measurement algorithm uses 3D cameras and other hardware to obtain its volume data passively,which produce high cost.This dissertation proposes a vol-ume measurement algorithm based on the stacking of multiple regular objects in a fixed scene,which can quickly and accurately obtain its volume data in logistics warehouses and other venues.According to the characteristics of the fixed scene and the imaging principle of the camera,this dissertation uses a simple geometric relationship computation to remove the redundant part of the external cube of the stacked object and obtain the approximate volume of the stacked object.Combined with the volume calculated by the 3D reconstruction,the accurate volume data can be calculated out using the weight ratio.Experimental results show that the volume measurement error of the algorithm reaches 4.56%,which meets the needs of actual engineering.3.Considering the convenience of the courier,this dissertation designs an App to simulate the courier to measure the volume of a single regular object.After actual testing,it can effectively prove the practicality and accuracy of the algorithm.