Detect material volume by fusing heterogeneous camera target detection and depth estimation information
Material piles, such as coal piles, sand piles, and gravel piles, are ubiquitous in real life, but in actual production, it is not easy to measure the remaining volume. The visual measurement algorithm is the mainstream solution for the volume measurement. Monocular measurement is fast, but there ar...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
AIP Publishing LLC
2025-01-01
|
| Series: | AIP Advances |
| Online Access: | http://dx.doi.org/10.1063/5.0246825 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832542738370789376 |
|---|---|
| author | Wei Tian Xuecong Cheng Yipeng Zhang Huazhi Lin |
| author_facet | Wei Tian Xuecong Cheng Yipeng Zhang Huazhi Lin |
| author_sort | Wei Tian |
| collection | DOAJ |
| description | Material piles, such as coal piles, sand piles, and gravel piles, are ubiquitous in real life, but in actual production, it is not easy to measure the remaining volume. The visual measurement algorithm is the mainstream solution for the volume measurement. Monocular measurement is fast, but there are problems of low accuracy and instability; binocular measurement is better, but it is not easy to extract the volume of the measured object; multi-eye measurement is accurate but slow. To combine the advantages of monocular and binocular measurements, this paper proposes an algorithm that integrates heterogeneous camera depth estimation and target detection information to realize material volume detection. First, the improved DeepLabV3+ is used to detect the edge of the material pile in the monocular camera target detection, and the CREStereo cascade network is used in the binocular camera to calculate the depth map; then, SIFT is combined with FLANN to map the edge of the material pile into the depth map and separate the depth of the material pile; finally, the three-dimensional coordinates of each point in the material pile are calculated, and the volume is calculated using the microelement method. Experiments show that the average accuracy of this method is 92.9%. |
| format | Article |
| id | doaj-art-a8128c7fc82f4250922492aa60738266 |
| institution | Kabale University |
| issn | 2158-3226 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | AIP Publishing LLC |
| record_format | Article |
| series | AIP Advances |
| spelling | doaj-art-a8128c7fc82f4250922492aa607382662025-02-03T16:40:42ZengAIP Publishing LLCAIP Advances2158-32262025-01-01151015132015132-1810.1063/5.0246825Detect material volume by fusing heterogeneous camera target detection and depth estimation informationWei Tian0Xuecong Cheng1Yipeng Zhang2Huazhi Lin3Southeast University, Nanjing 210096, ChinaCCCC Second Harbor Engineering Company Ltd., Wuhan 430040, ChinaCCCC Second Harbor Engineering Company Ltd., Wuhan 430040, ChinaSchool of Automation, Wuhan University of Technology, Wuhan 430040, ChinaMaterial piles, such as coal piles, sand piles, and gravel piles, are ubiquitous in real life, but in actual production, it is not easy to measure the remaining volume. The visual measurement algorithm is the mainstream solution for the volume measurement. Monocular measurement is fast, but there are problems of low accuracy and instability; binocular measurement is better, but it is not easy to extract the volume of the measured object; multi-eye measurement is accurate but slow. To combine the advantages of monocular and binocular measurements, this paper proposes an algorithm that integrates heterogeneous camera depth estimation and target detection information to realize material volume detection. First, the improved DeepLabV3+ is used to detect the edge of the material pile in the monocular camera target detection, and the CREStereo cascade network is used in the binocular camera to calculate the depth map; then, SIFT is combined with FLANN to map the edge of the material pile into the depth map and separate the depth of the material pile; finally, the three-dimensional coordinates of each point in the material pile are calculated, and the volume is calculated using the microelement method. Experiments show that the average accuracy of this method is 92.9%.http://dx.doi.org/10.1063/5.0246825 |
| spellingShingle | Wei Tian Xuecong Cheng Yipeng Zhang Huazhi Lin Detect material volume by fusing heterogeneous camera target detection and depth estimation information AIP Advances |
| title | Detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| title_full | Detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| title_fullStr | Detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| title_full_unstemmed | Detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| title_short | Detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| title_sort | detect material volume by fusing heterogeneous camera target detection and depth estimation information |
| url | http://dx.doi.org/10.1063/5.0246825 |
| work_keys_str_mv | AT weitian detectmaterialvolumebyfusingheterogeneouscameratargetdetectionanddepthestimationinformation AT xuecongcheng detectmaterialvolumebyfusingheterogeneouscameratargetdetectionanddepthestimationinformation AT yipengzhang detectmaterialvolumebyfusingheterogeneouscameratargetdetectionanddepthestimationinformation AT huazhilin detectmaterialvolumebyfusingheterogeneouscameratargetdetectionanddepthestimationinformation |