Computational Intelligence and Neuroscience

Volume 2018, Article ID 7690435, 7 pages

https://doi.org/10.1155/2018/7690435

## A Water-Area Recognition Approach Based on “Tuned” Texture Mask and Cuckoo Search Algorithm

^{1}State Grid Hunan Electric Power Corporation Economy Institute, Changsha 410003, China^{2}State Grid Hunan Electric Power Company Limited, Changsha 410003, China

Correspondence should be addressed to Kangbo Peng; moc.361@bkpstyd

Received 17 May 2018; Revised 9 August 2018; Accepted 16 October 2018; Published 9 December 2018

Academic Editor: Amparo Alonso-Betanzos

Copyright © 2018 Zhongwei Chen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

Texture feature extraction is a key topic in many applications of image analysis; a lot of techniques have been proposed to measure the characteristics of this field. Among them, texture energy extracted with a mask is a rotation and scale invariant texture descriptor. However, the tuning process is computationally intensive and easily trap into the local optimum. In the proposed approach, a “Tuned” mask is utilized to extract water and nonwater texture; the optimal “Tuned” mask is acquired by maximizing the texture energy value via a newly proposed cuckoo search (CS) algorithm. Experimental results on samples and images show that the proposed method is suitable for texture feature extraction, the recognition accuracy is higher than the genetic algorithm (GA), particle swarm optimization (PSO) and the gravitational search algorithm (GSA) optimized “Tuned” mask scheme, and the water area could be well recognized from the original image. Experimental results show that the proposed method could exhibit better performance than other methods involved in the paper in terms of optimization ability and recognition result.

#### 1. Introduction

Texture [1, 2] is a core property of the appearance of objects in natural scenes and is a powerful visual cue, used by both humans and machines in describing and recognizing objects of the real world. Texture feature extraction [3, 4] is a vital topic in machine vision and image analysis, which is to identify a texture sample as one of the several possible classes with a reliable texture classifier, and plays a very important role in a wide range of applications. There are kinds of texture due to changes in orientation, scale, or other visual appearances; as a result, a lot of texture feature extraction methods, such as grey-level co-occurrence matrix (GLCM), local binary pattern (LBP), Gabor wavelet, fractal theory, run-length texture descriptor, and so on [5–10], have been proposed over the years. More importantly, texture feature is one of the most significant symbols for different objects in remote sensing image, and water area is one of the most important landscape elements. Extraction of water area with image has become the favored technique to monitor urban expansion and environment, which is significant to the regional sustainable development. Some studies have been focused on the field of water area recognition by texture feature. Mettes et al. [11] proposed a hybrid texture descriptor and local detection algorithm to detect water in videos, which was able to robustly detect a region of water with high detection accuracy. Deng et al. [12] presented a high precision object-oriented water extraction scheme based on GLCM and decomposition approach, which was able to distinguish the influence of other objects and detect the small water area. However, it expends numerous features to complete the task of texture feature extraction for some traditional techniques, which need a large amount of CPU time to extract the features, and the excessive features will decrease the recognition efficiency at the same time. Although there are some methods that only need a few of features, it is difficult to stably obtain high recognition accuracy.

In order to extract the texture feature efficiently and effectively, the texture feature extraction technique based on texture mask has drawn rather considerable interest [13]. Among them, Law’s mask [14] is one of the most commonly used masks to classify the different types of texture. However, the basic form of Law’s mask is stationary, which is difficult to adapt various types of texture for a fixed mask [15]. Thus, You and Cohen [16] developed an adaptive texture feature extraction method called “Tuned” mask exempted from changes in rotation and scale of the texture image, and its validity has been proved. To obtain the optimal texture mask, it utilized a search strategy of gradient estimation and random search with heuristic learning. However, it is easy to lead to the high time complexity and probably trap into the local optimum.

In essence, how to obtain the optimal “Tuned” mask is a combinatorial optimization problem which may be handled by swarm intelligence algorithms. For example, Zheng and Zheng [17] utilized the genetic algorithm (GA) to search for the optimal “Tuned” mask and produced rather good results than random search. Ye et al. [18] explained the principle and steps of “Tuned” mask with the particle swarm optimization algorithm (PSO) and illustrated how to train “Tuned” mask with the proposed method in details. Wan et al. [19] introduced a residential area recognition method based on “Tuned” mask and optimized with the gravitational search algorithm (GSA), which was able to keep a good balance on the efficiency and recognition accuracy. In all, GA, PSO, and GSA could obtain good “Tuned” mask, but the dimension of the optimization problem is relatively high and the value of each individual should be a real number in the range of wide continuous space, which may not guarantee the research ability in the solution space; it is worth trying more swarm intelligence algorithms on this topic.

Cuckoo search (CS) [20] is a newly proposed swarm intelligence algorithm with stochastic global search strategy. Nowadays, the CS algorithm has been widely used in diverse applications, e.g., Wang et al. [21] utilized the CS algorithm to solve the function optimization problem and attained the optimal solution. Fouladgar et al. [22] utilized the CS algorithm to create a precise equation for predicting the ground vibration produced by blasting operations in copper mine. Suresh et al. [23] used the CS algorithm to make contrast enhancement for satellite images. In the field of texture feature extraction, Wang et al. [24] hybridized the CS algorithm with K-means algorithm to optimize the clustering center and enhanced the accuracy and efficiency of classification. Yang et al. [25] presented a remote image classification approach by learning the attribute weight of Naive Bayes classifier through the CS algorithm, which obtained higher classification accuracy and more stable performance than other evolutionary algorithms. Further, Medjahed et al. [26] proposed a new procedure for band selection by using the binary-coded CS algorithm to optimize the objective function, which could obtain satisfactory results with regard to other relevant approaches. In the paper, how to obtain the optimal “Tuned” mask is a continuous combinatorial optimization problem, which could be solved by decimal encoding. Hence, a novel water area recognition technique is proposed using “Tuned” mask and blending of the CS algorithm.

The rest of this paper is structured as follows. The idea of the proposed approach to produce the optimal “Tuned” mask by using the CS algorithm is detailed in Section 2. Section 3 displays the experimental results and discussion. Finally, the paper is concluded in Section 4.

#### 2. The Proposed Method

##### 2.1. Cuckoo Search Algorithm

Cuckoo search (CS) is a novelty evolutionary algorithm proposed by Yang and Deb in 2009. The algorithm is a search strategy model on brood parasitism of some cuckoo species by laying their eggs in the nests of other host birds. If a host bird discovers the eggs are not their own, they will either throw these alien eggs away or simply abandon its nest and build a new nest elsewhere. The better new solution will take place of the solution which is relatively worse in the nest. For simplicity, only three idealized rules are used to describe the CS algorithm as follows [27]:(i)Each cuckoo lays one egg at a time and dumps it in a randomly chosen nest.(ii)The best nests with high quality of eggs (solutions) will carry over to the next generations.(iii)The number of available host nests is fixed, and a host can discover an alien egg with a probability . The host bird can either throw the egg away or abandon the nest so as to build a completely new nest in a new location.

Moreover, a mass of studies have indicated that flight behaviors of many animals and insects have the typical characteristics of the Lévy flight. For an optimization problem, the quality of a solution could simply be corresponding to the fitness value of the objective function. Other forms of fitness can be defined in a parallel way to the objective function in other evolutionary algorithms. Three rules are defined in the algorithm; first, each egg in a nest stands for a solution; second, a cuckoo egg denotes a new solution; third, all of the cuckoos are evaluated by the fitness value of the objective function to be optimized and have velocities which directly decide the cuckoos’ flying; the intent is to use the new better solutions to replace the not so good solution in the nests.

In order to generate the new solutions , call the bird *i*; a Lévy flight can be defined as follows:where is the step size which should be connected with the solve space. The product means entry-wise multiplications. This entry-wise product is similar to those used in PSO, the random walk via Lévy flight is more efficient in searching the solve space, and its step length is much longer in the long run.

The Lévy flight essentially provides a random walk while the random step length is drawn from a Lévy distribution, which has an infinite variance with an infinite mean as follows:

Here, the consecutive steps of a cuckoo essentially from a random walk process which obeys a power law step-length distribution with a heavy tail. However, a large proportion of the new solutions may be generated by extensive randomization whose locations may be far from the current best solution; this will make sure the algorithm will not fall into a local optimum.

##### 2.2. “Tuned” Mask

In order to utilize the optimal texture mask and make an accurate recognition for different texture features, You and Cohen [16] suggested the extension of Law’s scheme by abandoning the traditional masks with constants and replacing them with variables in order to improve the recognition accuracy and reliability. In the method, a single 5 × 5 mask is produced which extracts a common feature of a single texture at different rotations and scales; at the same time, it discriminates this feature from other texture features to a large extent. The new mask is called a “Tuned” or adaptive mask, and the whole process of texture feature extraction is very simple.

The key issue to apply the CS algorithm is the representation of the problem, that is, how to make a suitable mapping between the problem solution and each bird. In the paper, a search space for a mask is 25 dimensions. Each dimension with continuous values and the symmetrical mask with zero sums are utilized to avoid plenty of computation, and the whole mask could be composed by only 10 parameters [17]. Therefore, the “Tuned” mask could be defined as follows:

As the size of the “Tuned” mask is 5 × 5 and requires being symmetrical with zero sums, so only 10 parameters in a mask need to be encoded, the layout of parameters in the mask plays a more important role for texture image classification than its actual values. Because the decimal code can be directly used for CS algorithm, the parameters of is encoded in the range of [−50, 50] for simplicity.

The “texture energy” TE could be calculated by the variance statistic within macro-window size in the training stage, which is defined as follows:where is the image after transformation with the optimal “Tuned” mask at the pixel point , is the coding for the “Tuned” mask, and is a window at pixel point (9 × 9 is used in the paper).

##### 2.3. Implementation of the Proposed Method

The proposed method is simple to implement. The main process to learn the “Tuned” mask based on CS algorithm for water area recognition and texture feature extraction is in Pseudocode 1.

According to the operational process of swarm intelligence algorithms, the computational results mainly depend on parameters setting in some extent; fine tuning of the parameters can produce a better result. Table 1 shows the parameters used in CS algorithm.