Z2-ZNCC: ZigZag Scanning based Zero-means Normalized Cross Correlation for Fast and Accurate Stereo Matching on Embedded GPU

Chang, Qiong; Zha, Aolong; Wang, Weimin; Xin, Liu; Onishi, Masaki; Matsumoto, Tsutomu

doi:10.1109/iccd50377.2020.00104

Cited by 5 publications

(5 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As expected, the implementations which are optimized and deployed on FPGA architectures are superior to those running on an embedded GPU. Furthermore, the superiority in terms of throughput of algorithms, such as those from Cui and Dahnoun [ 41 ] and Chang et al [ 26 ] that do not rely on a complex regularization scheme such as the SGM is also to be expected. Nonetheless, our approach has a lower throughput than a similar implementation of Hernandez-Juarez et al [ 25 ], while simultaneously being deployed on a more powerful system.…”

Section: Resultsmentioning

confidence: 99%

—SGM configuration of our approach has a lower throughput than that of a comparable configuration by Hernandez-Juarez et al [ 25 ], while at the same time running on a hardware generation that is two times newer and that has twice the number of CUDA cores.…”

Section: Discussionmentioning

confidence: 99%

“…In their work, Hernandez-Juarez et al [ 25 ] have also deployed their implementation on the NVIDIA Jetson TX1, encapsulating the embedded Tegra X1 GPU, reaching 42 FPS (4.19 FPS/W) on VGA resolution and a four path SGM optimization. Chang et al [ 26 ] further optimized the computation of the normalized cross-correlation (NCC) matching cost for the use on GPUs and deployed their SGM-based stereo algorithm on the NVIDIA Jetson TX2 reaching 28 FPS on images with a size of 1242 × 375 pixels.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

ReS2tAC—UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

Ruf

Mohrs

Weinmann

et al. 2021

Sensors

View full text Add to dashboard Cite

With the emergence of low-cost robotic systems, such as *UAV, the importance of embedded high-performance image processing has increased. For a long time, FPGAs were the only processing hardware that were capable of high-performance computing, while at the same time preserving a low power consumption, essential for embedded systems. However, the recently increasing availability of embedded GPU-based systems, such as the NVIDIA Jetson series, comprised of an ARM CPU and a NVIDIA Tegra GPU, allows for massively parallel embedded computing on graphics hardware. With this in mind, we propose an approach for real-time embedded stereo processing on ARM and CUDA-enabled devices, which is based on the popular and widely used Semi-Global Matching algorithm. In this, we propose an optimization of the algorithm for embedded CUDA GPUs, by using massively parallel computing, as well as using the NEON intrinsics to optimize the algorithm for vectorized SIMD processing on embedded ARM CPUs. We have evaluated our approach with different configurations on two public stereo benchmark datasets to demonstrate that they can reach an error rate as low as 3.3%. Furthermore, our experiments show that the fastest configuration of our approach reaches up to 46 FPS on VGA image resolution. Finally, in a use-case specific qualitative evaluation, we have evaluated the power consumption of our approach and deployed it on the DJI Manifold 2-G attached to a DJI Matrix 210v2 RTK *UAV, demonstrating its suitability for real-time stereo processing onboard a *UAV.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

ReS2tAC—UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

Ruf

Mohrs

Weinmann

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Compared to the related work from literature, the throughput and processing speed of our approach are not very impressive. We expected to reach a significantly lower throughput than approaches running on FPGAs [17,18], as well as a slightly lower throughput compared to approaches that do not rely on a computationally expensive optimization scheme but run on an embedded GPU [26,41]. However, the CT 9×7 -SGM configuration of our approach has a lower throughput than that of a comparable configuration by Hernandez-Juarez et al [25], while at the same time running on a hardware generation that is two times newer and that has twice the number of CUDA cores.…”

Section: Processing Speedmentioning

confidence: 96%

“…As expected, the implementations which are optimized and deployed on FPGA architectures are superior to those running on an embedded GPU. Furthermore, the superiority in terms of throughput of algorithms, such as those from Cui and Dahnoun [41] and Chang et al [26], that do not rely on a complex regularization scheme like the SGM is also to be expected. Nonetheless, our approach has a lower throughput than a similar implementation of Hernandez-Juarez et al [25], while simultaneously being deployed on a more powerful system.…”

Section: Throughput Frame Rates and Power Consumptionmentioning

confidence: 99%

ReS2tAC -- UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

Ruf,

Mohrs,

Weinmann

et al. 2021

Preprint

View full text Add to dashboard Cite

With the emergence of low-cost robotic systems, such as unmanned aerial vehicle, the importance of embedded high-performance image processing increased more and more. For a long time, FPGAs were the only processing hardware, that were capable of high-performance computing, while at the same time preserving a low power consumption, essential for embedded systems. However, the recently increasing availability of embedded GPU-based systems, such as the NVIDIA Jetson series, comprised of an ARM CPU and a NVIDIA Tegra GPU, allows for massively parallel embedded computing on graphics hardware. With this in mind, we propose an approach for real-time embedded stereo processing on ARM and CUDA enabled devices, which is based on the popular and widely used Semi-Global Matching algorithm. In this, we propose an optimization of the algorithm for embedded CUDA GPUs, by utilizing massively parallel computing, as well as utilizing the NEON intrinsics to optimize the algorithm for vectorized SIMD processing on embedded ARM CPUs. We have evaluated our approach with different configurations on two public stereo benchmark datasets to demonstrate that they can reach an error rate as low as 3.3 %. Furthermore, our experiments show that the fastest configuration of our approach reaches up to 46 FPS on VGA image resolution. Finally, in a use-case specific qualitative evaluation, we have evaluated the power consumption of our approach and deployed it on the DJI Manifold 2-G attached to a DJI Matrice 210v2 RTK unmanned aerial vehicle (UAV), demonstrating its suitability for real-time stereo processing onboard a UAV.

show abstract

Efficient stereo matching on embedded GPUs with zero-means cross correlation

Chang

Zha

Wang

et al. 2022

Journal of Systems Architecture

View full text Add to dashboard Cite

Z2-ZNCC: ZigZag Scanning based Zero-means Normalized Cross Correlation for Fast and Accurate Stereo Matching on Embedded GPU

Cited by 5 publications

References 8 publications

ReS2tAC—UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

ReS2tAC—UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

ReS2tAC -- UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices

Efficient stereo matching on embedded GPUs with zero-means cross correlation

Contact Info

Product

Resources

About