Difference between revisions of "Hyperspectral Imaging"
| Line 13: | Line 13: | ||
Hyperspectral imaging systems capture information in multiple wavelength bands across the electromagnetic spectrum. These bands provide substantial details based on the optical properties of the materials present in the captured scene. The high cost of hyperspectral cameras and their strict illumination requirements make the technology out of reach for end-user and small-scale commercial applications. We propose MobiSpectral, which turns a low-cost phone into a simple hyperspectral imaging system, without any changes in the hardware. We design deep learning models that take regular RGB images and near-infrared (NIR) signals (which are used for face identification on recent phones) and reconstruct multiple hyperspectral bands in the visible and NIR ranges of the spectrum. Our experimental results show that MobiSpectral produces accurate bands that are comparable to ones captured by actual hyperspectral cameras. The availability of hyperspectral bands that reveal hidden information enables the development of novel mobile applications that are not currently possible. To demonstrate the potential of MobiSpectral, we use it to identify organic solid foods, which is a challenging food fraud problem that is currently partially addressed by laborious, unscalable, and expensive processes. We collect large datasets in real environments under diverse illumination conditions to evaluate MobiSpectral. Our results show that MobiSpectral can identify organic foods, e.g., apples, tomatoes, kiwis, strawberries, and blueberries, with an accuracy of up to 94% from images taken by phones. | Hyperspectral imaging systems capture information in multiple wavelength bands across the electromagnetic spectrum. These bands provide substantial details based on the optical properties of the materials present in the captured scene. The high cost of hyperspectral cameras and their strict illumination requirements make the technology out of reach for end-user and small-scale commercial applications. We propose MobiSpectral, which turns a low-cost phone into a simple hyperspectral imaging system, without any changes in the hardware. We design deep learning models that take regular RGB images and near-infrared (NIR) signals (which are used for face identification on recent phones) and reconstruct multiple hyperspectral bands in the visible and NIR ranges of the spectrum. Our experimental results show that MobiSpectral produces accurate bands that are comparable to ones captured by actual hyperspectral cameras. The availability of hyperspectral bands that reveal hidden information enables the development of novel mobile applications that are not currently possible. To demonstrate the potential of MobiSpectral, we use it to identify organic solid foods, which is a challenging food fraud problem that is currently partially addressed by laborious, unscalable, and expensive processes. We collect large datasets in real environments under diverse illumination conditions to evaluate MobiSpectral. Our results show that MobiSpectral can identify organic foods, e.g., apples, tomatoes, kiwis, strawberries, and blueberries, with an accuracy of up to 94% from images taken by phones. | ||
| − | [[File:Mobispectral.png|thumb|center|700px|MobiSpectral | + | [[File:Mobispectral.png|thumb|center|700px|Overview of MobiSpectral]] |
== Code and Datasets == | == Code and Datasets == | ||
Revision as of 14:14, 21 September 2023
A hyperspectral camera captures a scene in many frequency bands across the spectrum, providing rich information and facilitating numerous applications in industrial, military, and commercial domains. Example applications of hyperspectral imaging include medical diagnosis (e.g., early detection of skin cancer), food-quality inspection, artwork authentication, forest monitoring, material identification, and remote sensing. The potential of hyperspectral imaging has been established for decades now. However, to date, hyperspectral imaging has only seen success in specialized and large-scale industrial and military applications. This is because of three main challenges: (i) the sheer volume of hyperspectral data which makes it hard to transmit such data in real-time and thus limiting its usefulness for many applications, (ii) the negative impact of the environmental conditions (e.g., rain, fog, and snow) which reduces the utility of the captured hyper-spectral data, and (iii) the high cost of hyperspectral cameras (upwards of $20K USD) which makes the technology out of reach for many commercial and end-user applications. The goal of this project is to address these challenges to enable wide adoption of hyperspectral imaging in many applications.
People
- Neha Sharma
- Puria Azadi
- Mohammad Amin Arab
- Kiana Calagari
- Mohamed Hefeeda
MobiSpectral: Hyperspectral Imaging on Mobile Devices
Hyperspectral imaging systems capture information in multiple wavelength bands across the electromagnetic spectrum. These bands provide substantial details based on the optical properties of the materials present in the captured scene. The high cost of hyperspectral cameras and their strict illumination requirements make the technology out of reach for end-user and small-scale commercial applications. We propose MobiSpectral, which turns a low-cost phone into a simple hyperspectral imaging system, without any changes in the hardware. We design deep learning models that take regular RGB images and near-infrared (NIR) signals (which are used for face identification on recent phones) and reconstruct multiple hyperspectral bands in the visible and NIR ranges of the spectrum. Our experimental results show that MobiSpectral produces accurate bands that are comparable to ones captured by actual hyperspectral cameras. The availability of hyperspectral bands that reveal hidden information enables the development of novel mobile applications that are not currently possible. To demonstrate the potential of MobiSpectral, we use it to identify organic solid foods, which is a challenging food fraud problem that is currently partially addressed by laborious, unscalable, and expensive processes. We collect large datasets in real environments under diverse illumination conditions to evaluate MobiSpectral. Our results show that MobiSpectral can identify organic foods, e.g., apples, tomatoes, kiwis, strawberries, and blueberries, with an accuracy of up to 94% from images taken by phones.
Code and Datasets
- MobiSpectral: Hyperspectral Imaging on Mobile Devices
- Enabling Hyperspectral Imaging in Diverse Illumination Conditions for Indoor Applications
- Hyperspectral Reconstruction from RGB Images for Vein Visualization
- Band and Quality Selection for Efficient Transmission of Hyperspectral Images
Publications
- N. Sharma, M. Waseem, S. Mirzaei, and M. Hefeeda, MobiSpectral: Hyperspectral Imaging on Mobile Devices, In Proc. of ACM Conference on Mobile Computing and Networking (MobiCom'23), Madrid, Spain, October 2023. (Artifacts Evaluated and Functional)
- P. Moghadam, N. Sharma, M. Hefeeda, Enabling Hyperspectral Imaging in Diverse Illumination Conditions for Indoor Applications, In Proc. of ACM Multimedia Systems Conference (MMSys’21), Istanbul, Turkey, September 2021. (Artifacts Evaluated and Functional)
- N. Sharma, M. Hefeeda, Hyperspectral Reconstruction from RGB Images for Vein Visualization, In Proc. of ACM Multimedia Systems Conference (MMSys'20), Istanbul, Turkey, June 2020. (Artifacts Evaluated and Functional)
- M. Arab, K. Calagari, M. Hefeeda, Band and Quality Selection for Efficient Transmission of Hyperspectral Images, In Proc. of ACM Conference on Multimedia (MM'19), Nice, France. October 2019.
