A hyperspectral image contains hundreds of narrow and contiguous spectral bands. Therefore, such images provide valuable information from the earth surface objects. Target detection (TD) is a fast growing research filed in processing hyperspectral images. In recent years, developing target detection algorithms has received growing interest in hyperspectral images. The aim of TD algorithms is to find specific targets with known spectral signatures. Nevertheless, the enormous amount of information provided by hyperspectral images increases the computational burden as well as the correlation among spectral bands. Besides, even the best TD algorithms exhibit a large number of false alarms due to spectral similarity between the target and background especially at subpixel level in which the size of target of interest is smaller than the ground pixel size of the image. Thus, dimensionality reduction is often conducted as one of the most important steps before TD to both maximize the detection performance and minimize the computational burden. However, in hyperspectral image analysis few studies have been carried out on dimension reduction or band selection for target detection in comparison to the hyperspectral image classification field. Otherwise band selection has great impact on remote sensing processing because of its effect on dimension reduction and reducing computational burden of hyperspectral image processing by selecting of optimum bands subset.

This paper presents a simple method to improve the efficiency of subpixel TD algorithms based on removing bad bands in a supervised manner. The idea behinds the proposed method is to compare field and laboratory spectra of desired target for detecting bad bands. Since the laboratory spectrum of targets is measured under standard conditions with the minimized level of noise and atmospheric effects, they can be considered as ideal spectrum. On the other hand, the recorded field-based reflectance spectrum are affected by surrounding objects such as vegetation cover and atmospheric affects specially water vapor absorption. Obviously, the spectrum becomes progressively noisier at longer wavelengths due to reduction of radiance of the illumination source, i.e., the sun. In this way, bad bands can be observed in the field based spectrum when comparing with the laboratory spectrum of the target of interest. Based on fitting a normal distribution to laboratory-field spectral difference of all corresponding bands, best of them will be select and introduce to target detection methods.

In this study for our evaluation, the proposed method is compared with six popular band selection methods implemented in PRtools and False alarm parameter for validation is used in this study. Comparison is done using two well-known sub-pixel TD algorithms, the adaptive coherence estimator (ACE) and the constrained energy minimization (CEM), in the target detection blind test dataset. This dataset includes two HyMap radiance and reflectance images of Cooke City in Montana, USA. The images are obtained by an airborne HyMap sensor which has 126 spectral bands and a ground sample distance of 3m. This dataset includes 10 sub-pixel targets located in an open grass region. Experimental results show that the proposed method improves the efficiency of the ACE and CEM comparing with other band selection methods used. Between of all target detection experiments only in 12 percent results destroyed. Moreover, high speed, simplicity, low computational burden, and time consuming are the advantages of the proposed method.