Search

In this work, a multi-exposure method is proposed to increase the dynamic range (DR) of hyperspectral imaging using an InGaAs-based short-wave infrared (SWIR) hyperspectral line camera. Spectral signatures of materials were captured for scenarios in which the DR of a scene was greater than the DR of a line camera. To demonstrate the problem and test the proposed multi-exposure method, plastic detection in food waste and polymer sorting were chosen as the test application cases. The DR of the hyperspectral camera and the test samples were calculated experimentally. A multi-exposure method is proposed to create high-dynamic-range (HDR) images of food waste and plastic samples. Using the proposed method, the DR of SWIR imaging was increased from 43 dB to 73 dB, with the lowest allowable signal-to-noise ratio (SNR) set to 20 dB. Principal Component Analysis (PCA) was performed on both HDR and non-HDR image data from each test case to prepare the training and testing data sets. Finally, two support vector machine (SVM) classifiers were trained for each test case to compare the classification performance of the proposed multi-exposure HDR method against the single-exposure non-HDR method. The HDR method was found to outperform the non-HDR method in both test cases, with the classification accuracies of 98% and 90% respectively, for the food waste classification, and with 95% and 35% for the polymer classification.

In this paper, we present a hyper-spectral imaging system and practical calibration procedure using a low-cost calibration reference made of polytetrafluoroethylene. The imaging system includes a hyperspectral camera and an active source of illumination with a variable spectral distribution of intensity. The calibration reference is used to measure the relative reflectance of any material surface independent of the spectral distribution of light and camera sensitivity. Winter road conditions are taken as a test application, and several spectral images of snow, icy asphalt, dry as-phalt, and wet asphalt were made at different exposure times using different illumination spectra. Graphs showing measured relative reflectance for different road conditions support the conclusion that measurements are independent of illumination. Principal component analysis of the acquired spectral data for road conditions shows well separated data clusters, demonstrating the system’s suitability for material classification.

This paper shows that deep learning can eliminate the superimposed twin-image noise in phase images of Gabor holographic setup. This is achieved by the conditional generative adversarial model (C-GAN), trained by input-output pairs of noisy phase images obtained from synthetic Gabor holography and the corresponding quantitative noise-free contrast-phase image obtained by the off-axis digital holography. To train the model, Gabor holograms are generated from digital off-axis holograms with spatial shifting of the real image and twin image in the frequency domain and then adding them with the DC term in the spatial domain. Finally, the digital propagation of the Gabor hologram with Fresnel approximation generates a super-imposed phase image for the C-GAN model input. Two models were trained: a human red blood cell model and an elliptical cancer cell model. Following the training, several quantitative analyses were conducted on the bio-chemical properties and similarity between actual noise-free phase images and the model output. Surprisingly, it is discovered that our model can recover other elliptical cell lines that were not observed during the training. Additionally, some misalignments can also be compensated with the trained model. Particularly, if the reconstruction distance is somewhat incorrect, this model can still retrieve in-focus images.