Journal of Advanced Materials in Engineering (Esteghlal)

---

Recently, image processing technique and robotic vision are widely applied in fault detection of industrial products as well as document reading. In order to compare the captured images from the target, it is necessary to prepare a perfect image, then matching should be applied. A preprocessing must therefore, be done to correct the samples’ and or camera’s movement which can occur during the capturing of images. The Radon Transform technique is applied in this study which is inherently invariant to any movement, such as dislocation and rotation which leads to scale changing. According to this technique, simple methods are proposed to determine the degree of movement. Results of computer simulation show the priority of the proposed method to other techniques. The accuracy of the proposed algorithm is less than 0.1 degree and is applicable to different segments such as texts, tables, drawings, …, which are prepared in different writing languages by different devices such as digital camera, scanner, fax, and printer. Keywords: Image processing, Image matching, Radon Transform, Skew detection documents, Computer application in industry.

---

Quality control of textile products is an important stage in textile industries. To this end, the conventional method in fault detection is human inspection. In the present work, Wavelet transform was applied on images of simple circular knitted fabrics to diagnose five regular defects. The results showed that the method applied was accurate and fast in addition to being capable of determining fault position and dimensions. Therefore, the Wavelet transform method is suitable for online fault detection

---

With progress in radar systems, a number of methods have been developed for signal processing and detection in radars. A number of modern radar signal processing methods use time-frequency transforms, especially the wavelet transform (WT) which is a well-known linear transform. The interference canceling is one of the most important applications of the wavelet transform. In Ad-hoc detection methods, the interference is firstly canceled and then a simple detector, like an energy detector, is used. Therefore, we have used wavelet-based approaches to cancel the interference and then an energy detector has been employed. In this paper, it is shown that in practical cases where the performance of matched filter or near-matched filter is degraded, wavelet-based methods are more efficient. Also, we have shown that for cases where targets with slow radial velocity or one close to blind velocity are removed by the MTI filter, wavelet-based denoising has a better performance.

---

The theory of distributed detection is receiving a lot of attention. A common assumption used in previous studies is the conditional independence of the observations. In this paper, the optimization of local decision rules for distributed detection networks with correlated observations is considered. We focus on presenting the detection theory for parallel distributed detection networks with fixed fusion rules to develop a numeric algorithm based on Neyman-Pearson criterion. Simulation results are presented to demonstrate the efficiency and convergence properties of the algorithm.

---

In this paper, two suboptimum detectors are proposed for coherent radar signal detection in K-distributed clutter. Assuming certain values for several initial moments of clutter amplitude, the characteristic function of the clutter amplitude is approximated by a limited series. Using the Pade approximation, it is then converted to a rational fraction. Thus, the pdf of the clutter amplitude is obtained as a sum of simple exponential functions. Using such a pdf, we develop the suboptimum detectors PGLR and PAALR, which are simplified forms of the GLR and AALR. Computer simulations show that the suggested detectors have appropriate performance compared to OLD, GLR and AALR detectors.

---

In this paper, radar detection based on Monte Carlo sampling is studied. Two detectors based on Importance Sampling are presented. In these detectors, called Particle Detector, the approximated likelihood ratio is calculated by Monte Carlo sampling. In the first detector, the unknown parameters are first estimated and are substituted in the likelihood ratio (like the GLRT method). In the second detector, the averaged likelihood ratio is calculated by integrating out the unknown parameters (like the AALR method). Thanks to the numerical nature of these methods, they can be applied to many detection problems which do not have analytical solutions. Simulation results show that both the proposed detectors and the GLRT have approximately the same performance in problems to which the GLRT can be applied. On the other hand, the proposed detectors can be used in many cases for which either no ML estimate of unknown parameters exists or their prior distribution is known.