Ajay K. Mandava

Sky Segmentation by Fusing Clustering with Neural Networks

Sky segmentation is an important task for many applications related to obstacle detection and path planning for autonomous air and ground vehicles. In this paper, we present a method for the automated sky segmentation by fusing K-means clustering and Neural Network (NN) classifications. The performance of the method has been tested on images taken by two Hazcams (ie., Hazard Avoidance Cameras) on NASA’s Mars rover. Our experimental results show high accuracy in determining the sky area. The effect of various parameters is demonstrated using Receiver Operating Characteristic (ROC) curves.

A CUDA Based Implementation of Locally-and Feature-Adaptive Diffusion Based Image Denoising Algorithm

In this paper we introduce a parallel implementation of locally-and feature-adaptive diffusion based (LFAD) method for image denoising using NVIDIA CUDA framework and graphics processing units (GPUs). LFAD is a novel method for removing additive white Gaussian (AWG) noise in images reported to yield high quality denoised images [1]. It approaches each image region separately and uses different number of nonlinear anisotropic diffusion iterations for each region to attain best peak signal to noise ratio (PSNR). The inverse difference moment (IDM) feature is embedded into a modified diffusion function. As the method has attained highest performance in the class of advanced diffusion based methods and it is competitive with all the state-of-the-art methods, however computationally intensive when executed on the general purpose CPU. To improve the performance, we implemented using the CUDA computational framework. In order to minimize GPU kernel access to the global memory, we use shared memory and the texture memory per multiprocessor. The performance of the GPU implementation of the LFAD has been tested on the standard benchmark images. We demonstrate that with a single NVIDIA Tesla C2050 GPU we can expedite the sequential CPU implementation in most cases from 13 to 20 times.

Radioscopic inspection of cargo containers with megavoltage energy barriers

To perform the inspection of cargo containers the radioscopic screening is performed by switching between 6 and 9 MeV of boundary energies as rapidly as 200 times per second and measuring the penetration levels in the contents of cargo. This technology facilitates the material identification via the analyses of the ratios of signals obtained at nominal and dual energies, which are 6 and 9 MeV, respectively. The techniques are developed for (a) visualizing the contents to produce an image suitable for fast inspection by a human operator, and for (b) prompting the custom personnel about the location of suspicious objects. Specifically, nuclear materials are of interest. The experiments have been conducted with Linatron K9 device designed by Varian Security and Inspection Products [1]. The capabilities are demonstrated for detection of objects of interests for the steel shields of 10 inches of thickness.

Advantages and Challenges of Radioscopic Detection of Nuclear Materials in Cargo Containers with Two Megavoltage Energy Barriers

Abstract Megavoltage X-ray technology is utilized to detect fissile materials that can be smuggled by terrorists among commercial goods in cargo containers. Material discrimination with dual energy barriers is based on a ratio of penetration levels at respective energies. However, for a broad bremsstrahlung spectrum, the approach is not reliable because of its sensitivity to mass thickness. Furthermore, cargo containers usually have combinations of materials in a stack that further complicates material identification. It is imperative to study the capability of dual mega-electron-volt energy radioscopy to detect materials of interest for its practical application at customs. The time to perform this inspection automatically and the need to manually open the container for examination are to be minimized for the smooth transport of goods through the national border. In this work, Linatron K9, developed and manufactured by Varian Inc., Inspection and Security Products, is used for experimentation. By switching 6- and 9-MeV beams, an interlaced penetration response is obtained. The automated detection of materials of high atomic numbers in the stack of materials is performed by proposed adaptive thresholding algorithm. The evaluation of the system based on a worst case scenario shows that the system meets requirements defined in the congressional report in terms of true and false positive identification rates, smallest object resolution, and the processing time.

Unsupervised segmentation of soil x-ray microtomography images

Advances in X-ray microtomography (XMT) are opening new opportunities for examining soil structural properties and fluid distribution around living roots in-situ. The low contrast between moist soil, root and air-filled pores in XMT images presents a problem with respect to image segmentation. In this paper, we develop an unsupervised method for segmenting XMT images to pores (air and water), soil, and root regions. A feature-based segmentation method is provided to isolate regions that consist of similar texture patterns from an image based on the normalized inverse difference moment of gray-level co-occurrence matrix. The results obtained show that the combination of features, clustering, and post-processing techniques has advantageous over other advanced segmentation methods.

Adaptive non-linear diffusion in wavelet domain

Traditional diffusivity based denoising models detect edges by the gradients of intensities, and thus are easily affected by noise. In this paper, we develop a nonlinear diffusion denoising method which adapts to the local context and thus preserves edges and diffuses more in the smooth regions. In the proposed diffusion model, the modulus of gradient in a diffusivity function is substituted by the modulus of a wavelet detail coefficient and the diffusion of wavelet coefficients is performed based on the local context. The local context is derived directly by analyzing the energy of transform across the scales and thus it performs efficiently in the real-time. The redundant representation of the stationary wavelet transform (SWT) and its shift-invariance lend themselves to edge detection and denoising applications. The proposed stationary wavelet context-based diffusivity (SWCD) model produces a better quality image compared to that attained by two high performance diffusion models, i.e. higher Peak Signal-to-Noise Ratio on average and lesser artifacts and blur are observed in a number of images representing texture, strong edges and smooth backgrounds.