Christian Merfort

Target recognitions in multiple-camera closed-circuit television using color constancy

Abstract. People tracking in crowded scenes from closed-circuit television (CCTV) footage has been a popular and challenging task in computer vision. Due to the limited spatial resolution in the CCTV footage, the color of people’s dress may offer an alternative feature for their recognition and tracking. However, there are many factors, such as variable illumination conditions, viewing angles, and camera calibration, that may induce illusive modification of intrinsic color signatures of the target. Our objective is to recognize and track targets in multiple camera views using color as the detection feature, and to understand if a color constancy (CC) approach may help to reduce these color illusions due to illumination and camera artifacts and thereby improve target recognition performance. We have tested a number of CC algorithms using various color descriptors to assess the efficiency of target recognition from a real multicamera Imagery Library for Intelligent Detection Systems (i-LIDS) data set. Various classifiers have been used for target detection, and the figure of merit to assess the efficiency of target recognition is achieved through the area under the receiver operating characteristics (AUROC). We have proposed two modifications of luminance-based CC algorithms: one with a color transfer mechanism and the other using a pixel-wise sigmoid function for an adaptive dynamic range compression, a method termed enhanced luminance reflectance CC (ELRCC). We found that both algorithms improve the efficiency of target recognitions substantially better than that of the raw data without CC treatment, and in some cases the ELRCC improves target tracking by over 100% within the AUROC assessment metric. The performance of the ELRCC has been assessed over 10 selected targets from three different camera views of the i-LIDS footage, and the averaged target recognition efficiency over all these targets is found to be improved by about 54% in AUROC after the data are processed by the proposed ELRCC algorithm. This amount of improvement represents a reduction of probability of false alarm by about a factor of 5 at the probability of detection of 0.5. Our study concerns mainly the detection of colored targets; and issues for the recognition of white or gray targets will be addressed in a forthcoming study.

FIB-assisted a-SiGe:H/a-SiC:H alloy analysis for ultra-low biased multispectral pi x n sensors with enhanced color separation features and low reflective ZnO:Al back-contacts

Common security CCD and CMOS imaging systems are not able to distinguish colorimetrically between dangerous chemical substances, for example whitish powders [1]. Hydrogenated amorphous silicon (a-Si:H) with profiled bandgaps can be found in solar cells to optimize the collection of incoming photons [2]. We developed multicolor photodiodes based on a-Si:H with different spectral response characteristics for a reliable, fast, cheap and non-destructive identification of potentially dangerous substances. Optical and I-V measurements were performed to explore the effect of combining linearly graded a-SiC:H-/a-SiGe:H layers with low reflective aluminum doped zinc oxide (ZnO:Al) cathodes. We determined absorption coefficients and mobility-lifetime products (μτ) of graded and non-graded absorbers to calculate the penetration depth of photons at different energies into the device structure. This set of parameters enables an optimization of the intrinsic layers so that charge accumulations are generated precisely at defined device depths. Significant color separation improvements could be achieved by using ZnO:Al cathodes instead of commonly used ZnO:Al/Chromium (Cr) reflectors. As a result, we obtained multicolor diodes with highly precise adjustment of the spectral sensitivity ranging from 420 nm to 580 nm, reduced interference fringes and a very low reverse bias voltage of -2.5 V maximum. Three terminal device architectures with similar absorbers exhibit a shift from 440 nm to 630 nm by applying reverse voltages of, for instance, -11.5 V at 580 nm [3]. Present research efforts concentrate on further improvements of the absorption region to reduce the bias without affecting the optical sensor performance, using extensive bandgap engineering techniques.

The top down design flow of a-Si:H photodiodes with multivariate methods of analysis

A fast and reliable detection of potentially dangerous substances has become very important in ensuring civilian security. Currently, modern security systems have proven to be more effective on the basis that objects should be properly characterized and identified. For instance, chemical tests are used to identify samples of whitish powder that is suspected to be dangerous or illegal. Although these chemical tests are conducted very quickly, they are relatively expensive. However, well established methods of optical characterization offer a suitable alternative. The demand for low-cost and disposable devices have escalated the development of intelligent photodiodes, especially of tunable a-Si:H multispectral photodiodes1. Our aim of reengineering is to develop the best match for the spectral response adjustment. Unfortunately, it is not sufficient to optimize the spectral response only. The top down design flow begins with the calculation of the photocurrent for different combinations of light sources, spectral responses and whitish powder samples to build up a multivariate data set. The optimum combination is found at the point of intersection in the factor values in a 2-D scattergram. It is therefore, required that the use optimized photodiodes would simplify and accelerate the identification of potentially dangerous substances.

Proper spectral band adjustment for coloristic feature-based recognition of safety signs

The development of multispectral and hyperspectral systems in the area of civil security have created new opportunities with regard to mobile sampling as well as rapid detection for on-site analysis. The latest developments, especially in the area of optical detectors, and the constant improvements to microprocessor computing capacity, have had an enormous influence on coloristic analysis methods. Ongoing optimization of multispectral systems is leading to a vast quantity of new application scenarios due to the simplification of measuring systems and the independence of specialized lab environments. The objective is to adapt the developed algorithm1 to a selection of suitable spectral bands, from monochrome powder specimens to panchromatic safety signs.

Application specific band selection with multivariate methods of analysis for a-Si:H multispectral photodiodes

In the recent past the generation and processing of multispectral data have had an immense impact on optical characterization systems. A virtual test environment is used to examine which bands provide a high information density. The photocurrent j = ∫ E(λ)*Sabs(λ)*r(λ) dλ was calculated for different light sources E, spectral response curves Sabs (bands), and the reflectance r of whitish powder samples that were suspected to be dangerous or illegal. The multivariate dataset will have to be determined whether we can gain any knowledge from this. The employed factor analysis is a common method of the group of structure-discovering methods and provides good results in the discovery of connections between parameters. It is particularly used if a variety of parameters must be reduced for some reason. For the verification, a dimension of the external separation is defined. To carry out this an n-dimensional vector P must be assigned to each measurement that is registered in the matrix M to determine the volume V of this dot cloud. The dimension normalized volume is defined as ΔCL, where n is the quantity of employed bands. The reliability of the complete measurement system is made by a membership function μ(P) comparable to the definitions from the area of the fuzzy sets. The parameter μ indicates with which reliability a measured pattern P could be assigned to a sample S from a dataset. The use of such optimized multispectral photodiodes would simplify and accelerate the identification of potentially dangerous substances.