Aerial Image Stitching Algorithm for Unmanned Aerial Vehicles Based on Improved ORB and PROSAC

Zhenyu Li; Yuan Tian; Fangjie Chen; Jun Han

doi:10.3788/LOP56.231003

Laser & Optoelectronics Progress, Volume. 56, Issue 23, 231003(2019)

Aerial Image Stitching Algorithm for Unmanned Aerial Vehicles Based on Improved ORB and PROSAC

Zhenyu Li^1,2, Yuan Tian³, Fangjie Chen^4、*, and Jun Han⁴

¹State Grid Power System Artificial Intelligence Joint Lab, State Grid Shandong Electric Power Company Electric Power Research Institute, Jinan, Shandong 250001, China

²Shandong Luneng Intelligent Technology Co., Ltd., Jinan, Shandong 250002, China

³State Grid Shandong Electric Power Company, Jinan, Shandong 250001, China

⁴School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Determination of feature point main direction

Download full size

Fig. 2. Gradual-in and gradual-out image fusion method

Download full size

Fig. 3. Oxford VGG dataset. (a) Bikes; (b) boat; (c) graffiti

Download full size

Fig. 4. UAV aerial test images. (a) Test image 1 (b) test image 2; (c) test image 3

Download full size

Fig. 5. Comparison of stitching results of three groups of test images. (a) Stitching image 1; (b) stitching image 2; (c) stitching image 3

Download full size

Table 1. Average time of each algorithm for feature point extraction

View table

Table 1. Average time of each algorithm for feature point extraction

Algorithm	Time /ms
Algorithm	Fig. 3(a)	Fig. 3(b)	Fig. 3(c)	Fig. 4(a)	Fig. 4(b)	Fig. 4(c)
SIFTSURFORBProposed	1.8750.4860.0790.186	1.3960.3550.0580.137	0.9230.2410.0420.103	7.6642.2750.4220.821	8.3212.0310.3870.933	7.8532.1320.4380.877

Table 2. Matching accuracy of each algorithm

View table

Table 2. Matching accuracy of each algorithm

Algorithm	R_CM /%
Algorithm	Fig. 3(a)	Fig. 3(b)	Fig. 3(c)	Fig. 4(a)	Fig. 4(b)	Fig. 4(c)
SIFT+RANSACSURF+RANSACORB+RANSACORB+PROSACProposed	97.3596.2592.3795.0396.58	96.7994.8390.3193.7696.27	95.2895.0490.2193.4495.53	98.1296.6791.7394.3997.53	97.2395.4892.4795.2996.79	96.4695.7989.8693.5396.21

Table 3. Matching time of each algorithm

View table

Table 3. Matching time of each algorithm

Algorithm	Time /ms
Algorithm	Fig. 3(a)	Fig. 3(b)	Fig. 3(c)	Fig. 4(a)	Fig. 4(b)	Fig. 4(c)
SIFT+RANSAC	112.88	115.34	120.74	133.55	154.79	125.56
SURF+RANSAC	102.53	97.64	104.27	115.57	123.26	99.74
ORB+RANSAC	93.79	90.56	93.85	92.49	117.38	103.19
ORB+PROSAC	51.69	50.73	51.22	58.35	64.28	55.86
Proposed	53.87	56.42	55.79	64.39	73.54	60.83

Table 4. Root mean square error of each algorithm
View table
Table 4. Root mean square error of each algorithm
Algorithm E_RMS
Fig. 4(a) Fig. 4(b) Fig. 4(c)
SIFT+RANSACSURF+RANSACORB+RANSACORB+PROSACProposed 0.4790.5450.7170.6130.529 0.5450.6830.7410.6620.567 0.5270.6030.6920.6440.586

Tools

Get Citation

Copy Citation Text

Zhenyu Li, Yuan Tian, Fangjie Chen, Jun Han. Aerial Image Stitching Algorithm for Unmanned Aerial Vehicles Based on Improved ORB and PROSAC[J]. Laser & Optoelectronics Progress, 2019, 56(23): 231003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites