Food recognition using neural network classifier and multiple hypotheses image segmentation

ABSTRACT

This paper proposes the multiple-hypotheses image segmentation and feed-forward neural network classifier for food recognition to improve the performance. Initially, the food or meal image is given as input. Then, the segmentation is applied to identify the regions, where a particular food item is located using salient region detection, multi-scale segmentation, and fast rejection. Then, the features of every food item are extracted by the global feature and local feature extraction. After the features are obtained, the classification is performed for each segmented region using a feed-forward neural network model. Finally, the calorie value is computed with the aid of (i) food volume and (ii) calorie and nutrition measure based on mass value. The experimental results and performance evaluation are validated. The outcome of the proposed method attains 0.947 for Macro Average Accuracy (MAA) and 0.959 for Standard Accuracy (SA), which provides better classification performance.     

Demonstration of a tool for automatic learning and re-use of knowledge in the activated sludge process.

Wastewater treatment plant operators encounter complex operational problems related to the activated sludge process and usually respond to these by applying their own intuition and by taking advantage of what they have learnt from past experiences of similar problems. However, previous process experiences are not easy to integrate in numerical control, and new tools must be developed to enable re-use of plant operating experience. The aim of this paper is to investigate the usefulness of a case-based reasoning (CBR) approach to apply learning and re-use of knowledge gained during past incidents to confront actual complex problems through the IWA/COST Benchmark protocol. A case study shows that the proposed CBR system achieves a significant improvement of the benchmark plant performance when facing a high-flow event disturbance.     

Benchmark simulation Model no 2 in Matlab-simulink: towards plant-wide WWTP control strategy evaluation.

In this paper, implementation of the Benchmark Simulation Model No 2 (BSM2) within Matlab-Simulink is presented. The BSM2 is developed for plant-wide WWTP control strategy evaluation on a long-term basis. It consists of a pre-treatment process, an activated sludge process and sludge treatment processes. Extended evaluation criteria are proposed for plant-wide control strategy assessment. Default open-loop and closed-loop strategies are also proposed to be used as references with which to compare other control strategies. Simulations indicate that the BM2 is an appropriate tool for plant-wide control strategy evaluation.     

WWTP dynamic disturbance modelling--an essential module for long-term benchmarking development.

Intensive use of the benchmark simulation model No. 1 (BSM1), a protocol for objective comparison of the effectiveness of control strategies in biological nitrogen removal activated sludge plants, has also revealed a number of limitations. Preliminary definitions of the long-term benchmark simulation model No. 1 (BSM1_LT) and the benchmark simulation model No. 2 (BSM2) have been made to extend BSM1 for evaluation of process monitoring methods and plant-wide control strategies, respectively. Influent-related disturbances for BSM1_LT/BSM2 are to be generated with a model, and this paper provides a general overview of the modelling methods used. Typical influent dynamic phenomena generated with the BSM1_LT/BSM2 influent disturbance model, including diurnal, weekend, seasonal and holiday effects, as well as rainfall, are illustrated with simulation results. As a result of the work described in this paper, a proposed influent model/file has been released to the benchmark developers for evaluation purposes. Pending this evaluation, a final BSM1_LT/BSM2 influent disturbance model definition is foreseen. Preliminary simulations with dynamic influent data generated by the influent disturbance model indicate that default BSM1 activated sludge plant control strategies will need extensions for BSM1_LT/BSM2 to efficiently handle 1 year of influent dynamics.     

Magnetotransport Calculations for Two-Dimensional Electrons with Spin-Orbit Interaction

We present calculations of the magnetoconductivity in a two-dimensional electron system including the Rashba spin-orbit interaction. Essential for these calculations is an extension of the self-consistent Born approximation which takes into account the electron spin degree of freedom. The calculated magnetoconductivity exhibits, besides the beating in the Shubnikov-de Haas oscillations, a modulation related to the spin-orbit induced crossings of Landau levels, as a consequence of spin-conserving scattering between spin-orbit coupled states.     

Anthocyanin composition of the red wine Babić affected by maceration treatment

Identification of anthocyanins in the wine made of the Croatian autochthonous grape variety of Babić (Vitis vinifera L.) was carried out and their profile was determined by means of high-performance liquid chromatography with diode array and mass spectrometric detection. Dependence of anthocyanins content and profile on maceration treatment conditions was investigated. Statistically significant differences of anthocyanins concentration in wines Babić produced by various maceration treatments were confirmed by the use of multivariate analysis of variances. The investigation results indicated that the maceration temperature exerts higher influence on anthocyanins concentration than the duration of maceration. In addition, on the basis of anthocyanin composition and using different multivariate statistical analyses, differentiation of wines Babić according to maceration treatments was procured.     

Alternating-direction implicit (ADI) formulation of the finite-difference time-domain (FDTD) method: algorithm and material dispersion implementation

The alternating-direction implicit finite-difference time-domain (ADI-FDTD) technique is an unconditionally stable time-domain numerical scheme, allowing the /spl Delta/t time step to be increased beyond the Courant-Friedrichs-Lewy limit. Execution time of a simulation is inversely proportional to /spl Delta/t, and as such, increasing /spl Delta/t results in a decrease of execution time. The ADI-FDTD technique greatly increases the utility of the FDTD technique for electromagnetic compatibility problems. Once the basics of the ADI-FDTD technique are presented and the differences of the relative accuracy of ADI-FDTD and standard FDTD are discussed, the problems that benefit greatly from ADI-FDTD are described. A discussion is given on the true time savings of applying the ADI-FDTD technique. The feasibility of using higher order spatial and temporal techniques with ADI-FDTD is presented. The incorporation of frequency dependent material properties (material dispersion) into ADI-FDTD is also presented. The material dispersion scheme is implemented into a one-dimensional and three-dimensional problem space. The scheme is shown to be both accurate and unconditionally stable.     

Time-resolved spectroscopy of infrared active vibrations in 2,5-dioctyloxy poly(phenylene vinylene) films

We report on picosecond time resolved spectroscopy of photogenerated infrared active vibrations in thin films of 2,5-dioctyloxy poly(phenylene vinylene). We excited the films by ?4 ps long pulses of 565 nm laser light with 2×10¹³ photons/cm² per pulse and repetition rate of 76 MHz. We then followed the temporal evolution of the infrared active vibrational (IRAV) spectrum using a subsequent, variably delayed, weak tunable IR probe pulses of similar temporal duration. Under these conditions, we show clear spectroscopic evidence for photogenerated infrared active vibrations at times which are shorter than our temporal resolution (<4 ps). We suggest that the transient IRAV absorption is due to secondary polarons formation following exciton dissociation.