Ali Can Karaca

Plastic waste sorting using infrared hyperspectral imaging system

The inherent chemical properties of materials can be brought into perspective using the large amount of spectral information provided by hyperspectral imaging systems. Therefore, the utilization of hyperspectral imaging in industrial applications is gradually increasing. One of the industrial sectors that can benefit from the advantages of hyperspectral imaging is recycling. Plastics which have different chemical properties (PP, PE, PVC, PET and PS) need sorting for plastic waste recycling. In this study, different types of plastics in hyperspectral images acquired using a shortwave infrared (SWIR) hyperspectral imaging system are successfully sorted.

Comparative evaluation of vector machine based hyperspectral classification methods

This paper presents a comparison of the classification performance of some vector machine based classification methods, namely, Import Vector Machines (IVM), Support Vector Machines (SVM) and Relevance Vector Machines (RVM), for hyperspectral images. Evaluation is carried out in terms of the number of vectors and classification accuracies. Furthermore, novel to this paper, Discriminative Random Field method with Graph Cut algorithm is applied to the probabilistic classification output of IVM based hyperspectral classification results, and it is shown that this approach significantly increases classification accuracies.

Local spectral gradient based gas detection in LWIR hyperspectral images

In this work, the use of local gradient features of spectral signatures with spectral angle mapper for target detection in long wave infrared hyperspectral images is proposed. The proposed method is tested for gas detection and experimental results show that utilizing local gradient features improve the performance.

Hyperspectral change detection with stereo disparity information enhancement

A hyperspectral change detection method with stereo depth information enhancement is proposed in this paper. The method operates on the hyperspectral data acquired by the ground-based hyperspectral stereo imaging system. The imaging system combines the properties of panoramic and stereo imaging with the high spectral resolution of hyperspectral cameras, and is of use especially for surveillance applications. Stereo and spectral information provided by the system are fused in the proposed method for enhanced change detection. Experimental results are evaluated on two hyperspectral datasets acquired by the system. Preliminary results show the improved performance of the system and the proposed method.

Hyperspectral change detection by multi-band Census Transform

Hyperspectral imaging provides increased capability for many image processing tasks with respect to standard imaging systems. One of such tasks in hyperspectral image processing is change detection, which aims to detect the differences occurring between images acquired from the same scene at different times. In this paper, a panoramic hyperspectral imaging system is used to capture multitemporal hyperspectral data, and novel multi-band Census Transform (MCT) is proposed for change detection on these data. Experimental results validate the performance of the proposed method for the utilized acquisition system.

Lossless hyperspectral image compression using bimodal conventional recursive least-squares

ABSTRACT Hyperspectral image compression is an important task, where the aim is to store or transmit data in an efficient way. Hyperspectral images are mostly captured by a sensor system that includes multiple imaging sensors covering different regions of the electromagnetic spectrum. Misalignment of multiple imaging sensors produces boresight effect, and this problem can degrade band prediction performance noticeably. Another problem is prediction of blurry band images. In order to gain robustness to these problems, bimodal conventional recursive least-squares (B-CRLS) prediction method is proposed for the lossless compression of hyperspectral images. Two prediction modes are defined: spectral and spatio-spectral. B-CRLS method has a two-step process. First, mode selection is carried out for each band. Afterwards, final band image is predicted by using the selected mode, and the residual images are encoded with an arithmetic encoder. The proposed method is compared to adaptive-length CRLS, fixed-length CRLS, and other well-known prediction methods. Experiments have been performed on uncalibrated and calibrated hyperspectral images. Obtained results show that the proposed method achieves competitive compression performance with the state-of-the-art while providing relatively lower-computation times.

Lossless compression of ultraspectral sounder data using recursive least squares

Lossless compression is an important topic in ultraspectral sounder data which includes thousands of spectral channels and it needs to store or transmit data in an efficient form. In this paper, a recursive least squares (RLS) based prediction method is proposed for the lossless compression of ultraspectral data. Experiments are performed on 10 granule maps which are acquired by NASAs Atmospheric Infrared Sounder (AIRS) system. The experimental results show that the proposed method provides comparable compression ratios to the-state-of-the-art-methods, i.e., ADQPCA and FSQPCA. Given its compression performance and lower complexity, the proposed method can be effectively implemented to embedded systems and it is well suited for onboard processing on satellites.

Comparison of traditional and recent unsupervised band selection approaches in hyperspectral images

In this paper, well-known traditional band selection methods which are used in hyperspectral imaging, namely, Maximum-Variance Principal Component Analysis (MVPCA), Maximum-SNR Principal Component Analysis (MSNRPCA), k-means, k-medoids, and recently proposed Automatic Band Selection (ABS) and Band Column Selection (BCS) approaches are compared. To assess the band selection performance of the methods, the change of classification performance by the number of selected bands is used. Performances of the methods are evaluated on three hyperspectral data sets and obtained results are compared in this paper.

A novel panoramic stereo hyperspectral imaging system

This paper proposes a novel panoramic stereo hyperspectral imaging system with theoretical analysis and experiments. The system has two line scan hyperspectral cameras and is set above a rotary stage. This provides the panoramic property of the system, and stereo hyperspectral panoramas are created at each 360 degree rotation. Furthermore, a novel algorithm is proposed, in which multi-band Census transform is implemented to estimate the disparity map information from the acquired stereo hyperspectral images. The system is utilizable for change detection, target detection and classification applications because it provides both disparity map and spectral property for any pixel or area in the image.

Target Preserving Hyperspectral Image Compression Using Weighted PCA and JPEG2000

Lossy compression methods can significantly reduce the volume of hyperspectral images. Besides that, target detection performance degrades dramatically at lower bit-rates. In this paper, we propose a target preserving compression method for low bit-rates. The proposed method consists of three parts. In the first part, a target detection algorithm is performed on hyperspectral image. Afterwards, a weight matrix is generated using output of the target detection. Finally, Weighted Principal Component Analysis (WPCA) and JPEG2000 methods are executed sequentially. Two different approaches are proposed for weight matrix generation and the proposed approaches are compared with PCA+JPEG2000 and SubPCA+JPEG2000 methods in terms of signal-to-noise ratio (SNR), receiver operating characteristic (ROC) curves and average mean square error. Experimental results demonstrate that WPCA+JPEG2000 provides significantly better target detection performance than other methods especially at low bit-rates.