Graphic object feature extraction system based on Cuckoo Search Algorithm

doi:10.1016/j.eswa.2016.08.068

Expert Systems with Applications

Volume 66, 30 December 2016, Pages 20-31

https://doi.org/10.1016/j.eswa.2016.08.068 Get rights and content

Highlights

•
Novel approach to image feature extraction using Bio Inspired Methods.
•
Simplified fitness condition to help in input image processing.
•
Simplified image filtering tailored for Bio Inspired Methods.
•
Solution works with or without ad-hoc filtering.
•
Dedicated Cuckoo Search Algorithm for graphic applications.

Abstract

Multimedia systems often use various detection methods to track relevant objects in images and video frames. The tracking scheme is often based on capturing of significant points in the object, which are used by implemented methods to extract the shape, dimensions, etc. and then further process these information. In recent years many advances in Computational Intelligence methods and approaches have been reported. Therefore the question has arisen if these, i.e. heuristics, are applicable to multimedia tracking systems?

This article is to discuss developed heuristic methods for Key-Points tracking and shape extraction. In the following sections of this work developed approaches, in particular a dedicated Cuckoo Search Algorithm and Firefly Algorithm versions, are presented and discussed in comparison to some classical methods to show potential advantages and disadvantages. Benchmark tests and experimental research results are presented to show efficacy and extraction precision on test images.

Introduction

Computer Science (CS) involves many aspects of Computational Intelligence (CI) to solve arising problems and develop solutions. CI methods and approaches very often simulate intelligence of nature in their applications in economic and industrial expert systems. One of these applications are in multimedia systems, where it is necessary to find something peculiar in the image or extract some features from the image for further processing.

Feature extraction is crucial for identification of users based on data streams. Cpalka, Zalasinski, and Rutkowski (2016) proposed new algorithm for dynamic handwritten signature analysis based on extraction of some typical aspects of lines that each person does while signing documents. Similarly these objects can be identified by devoted partitioning, where even on-line systems can efficiently extract information from signatures for verification purposes as presented by Cpalka and Zalasinski (2014). Lines and object Key-Points are also important for other extraction purposes, i.e. Meng and He (2016) presented consistent quadrangulation for shape feature line co-extraction, where devoted approach was implemented to assist in shape detection from images. However efficiency of this depends not only on the applied method but also on the image quality. Bera and Sychel (2016) described some improvement for object extraction from blurred images. Similarly hyperspectral images require sophisticated methods to process information and extract important features. Zabalza et al. (2016) presented devoted approach for these images where implemented auto encoder was processing segmented images. Other peculiar methods for feature extraction use i.e. dual-tree complex wavelet transforms as presented by Yang and Yang (2016), where gray-level co-occurrence matrix was used to calculate positions of the objects. However not all types of images can be processed like that. Multispectral Earth observation images present surface of our planet from the distance therefore to extract features it is necessary to apply multi-level systems where objects in the image will be processed for various aspects, this type of comparison is called patched-based classification, i.e. developed by Georgescu, Vaduva, Raducanu, and Datcu (2016). Sometimes classification methods can be tuned for faster processing. Korytkowski, Rutkowski, and Scherer (2016) proposed approach based on fuzzy classifiers to speed up the process of feature comparison. Acceleration of extraction has great importance for video-based tracking, where objects move and in case of low quality equipment the images can lose sharpness. Mithun, Howlader, and Rahman (2016) described methodology for vehicles tracking using multiple time-spatial images, which is devoted for video processing. Medicine is one of these fields of science and life, where all developments in automated expertise can find applications the fastest. In medicine various medical examinations are based on images, therefore developments in feature extraction and Key-Points search are very beneficial for medical systems. Gedik (2016) presented multi-resolution approach to extract important features from mammograms. Liu et al. (2016a) described methodology for identification of lesion images from gastrointestinal endoscope. Similarly pathology in internal organs of our bodies can be detected by extraction of information from images. Liu, Jiang, Xia, and Yi (2016b) proposed liver examinations based on multispatial mapping and statistical properties of images.

Among CI methods applied to extraction of important features from images heuristic approaches have been reported to show various advances in recent years. Walia and Kapoor (2014) developed intelligent video target tracking method using dedicated Cuckoo Search Algorithm, where heuristic method was used to filter the image from unnecessary information. Wozniak, Polap, Kosmider, Napoli, and Tramontana (2015) presented Cuckoo Search Algorithm (CSA) and Firefly Algorithm (FA) efficiency in medical images processing for extraction of degenerated tissues of lungs from x-ray images. Bhandari, Singh, Kumar, and Singh (2014) developed application of CSA in optimization of satellite image segmentation for multilevel thresholding, while Panda, Agrawal, and Bhuyan (2013) used CSA to edge magnitude. Mishra, Agarwal, Sharma, and Bedi (2014) presented these methods applied to gray-scale image watermarking over DWT SVD data. Heuristic approaches can also help in identification of humans as discussed by Rodrigues, Silva, Papa, Marana, and Yang (2016). Therefore the question has arisen if heuristics are also applicable to multimedia tracking systems?

This paper presents our experimental research results on application of devoted approach based on CSA and FA into a 2D image features extraction. Benchmark tests presented in this article are performed for sample images from open test databases.¹^,² Our results are compared to other conventional methods, the applied heuristic solution emerges as efficient in processing of various pictures. Performed experiments have low complexity, which makes the solution efficient, precise and easy to implement in comparison to classic methods.

Section snippets

2D Image Key-Points search methods

A computer image is a composition of points (pixels). Each pixel x_i in the image $I$ has measurable components of spatial coordinates x_{i, k}, where i is the number of the pixel and $k = 1, 2,$ as in the image $I$ there are two coordinates representing axes (see Fig. 1). In the proposed system, 2D images are input objects. Therefore points $x_{i} \in I$ are defined in two notations $x_{i} = (x, y)$ and $x_{i} = (x, y) = (x_{i, 1}, x_{i, 2}),$ as in the following sections both will be used to simplify the outlook of given equations.

Some

Novel approach to feature extraction

As presented in Section 2, a digital 2D image $I$ consists of points, each having a position and special properties. This composition brings unique information about objects in the picture and helps to recognize them. The proposed solution uses a special combination of image filtering with BIM processing to perform feature extraction for further recognition. Dedicated sobel filter is applied on the input image to be passed to the BIM based recognition. Filtering is used to extract the most

Feature extraction based on simplified sobel filter and Graphix Cuckoo Search Algorithm

In the proposed BIM solution, the algorithm presented in Section 3.2.1 is applied to search for Key-Points in sobel filtered images. Preprocessing is done using dedicated method for image filtering discussed in Section 3.1. Then, a population of cuckoos is simulated in order to move from pixel to pixel and search for Key-Points for feature extraction. This process is based on a simple fitness condition, where we want each of the extracted pixels to have at least 10 similarly recognized pixels

Experimental research results

In the experiments, sample test images from open data bases were used (see Section 1). Simulations were performed for GCSA using 500 cuckoos in only 10 generations with set coefficients: $β = 0.5,$ $γ = 0.3,$ $δ = 0.2,$ $μ = 0.25$ and $b e s t_r a t i o = 30 %$ . These coefficients represent the best settings for feature extraction among 100 benchmark samplings. It is hard to give the exact equations to precisely calculate these coefficients, therefore taking values from best Key-Points recognition among sampling benchmarks

Final remarks

In this article a novel approach to object detection and shape features extraction based on application of heuristic method, in particular developed Graphix Cuckoo Search Algorithm with a simplified sobel filter and devoted fitness condition, was discussed. Developed methodology can be implemented in various combinations, as track 1 without filtering and as tracks 2 and 3 with filtering. First approach is able to trace objects, while others because of applied filtering method in pre-processing

References (23)

A. Bhandari et al.
Cuckoo search algorithm and wind driven optimization based study of satellite image segmentation for multilevel thresholding using kapurs entropy
Expert Systems with Applications
(2014)
K. Cpalka et al.
On-line signature verification using vertical signature partitioning
Expert Systems with Applications
(2014)
K. Cpalka et al.
A new algorithm for identity verification based on the analysis of a handwritten dynamic signature
Applied Soft Computing
(2016)
N. Gedik
A new feature extraction method based on multi-resolution representations of mammograms
Applied Soft Computing
(2016)
M. Korytkowski et al.
Fast image classification by boosting fuzzy classifiers
Information Sciences
(2016)
LiuD. et al.
Identification of lesion images from gastrointestinal endoscope based on feature extraction of combinational methods with and without learning process
Medical Image Analysis
(2016)
MengM. et al.
Consistent quadrangulation for shape collections via feature line co-extraction
Computer-Aided Design
(2016)
A. Mishra et al.
Optimized gray-scale image watermarking using dwt svd and firefly algorithm
Expert Systems with Applications
(2014)
N.C. Mithun et al.
Video-based tracking of vehicles using multiple time-spatial images
Expert Systems with Applications
(2016)
R. Panda et al.
Edge magnitude based multilevel thresholding using cuckoo search technique
Expert Systems with Applications
(2013)

D. Rodrigues et al.

Eeg-based person identification through binary flower pollination algorithm

Expert Systems with Applications

(2016)

Cited by (40)

A knee-guided differential evolution algorithm for unmanned aerial vehicle path planning in disaster management
2021, Applied Soft Computing
Citation Excerpt :
As one branch of evolutionary algorithms (EAs), the cuckoo search algorithm (CSA) is a very efficient method, and in it, the Gaussian distribution is applied. It is regarded as a good candidate for the optimization problem of UAVs [42]. Table 1 lists recent and popular metaheuristics algorithms proposed to solve UAV path planning.
Unmanned aerial vehicles are instrumental in monitoring and analyzing information and searching for people in disaster relief scenarios. In this paper, path planning is constructed as a multiobjective optimization problem with constraints in a three-dimensional terrain disaster scenario. The objective functions involve the distance and risk of the path, which are calculated based on Bézier theory. The constraints include the turning angle and flight altitude. To solve this problem in an effective and efficient manner, a differential evolution algorithm that is based solely on the knee point is proposed, in which the knee solution would guide the search direction of the algorithm. According to the minimal Manhattan distance approach, the algorithm can quickly identify an optimal solution to generating a smooth path for decision-makers. Experimental results have confirmed the superiority of the proposed algorithm, and the rankings of the minimal Manhattan distance approach are consistent with multicriteria decision-making methods.
Design of optimal high pass and band stop FIR filters using adaptive Cuckoo search algorithm
2018, Engineering Applications of Artificial Intelligence
Citation Excerpt :
CSA can be efficiently used to an optimized multilevel thresholding problem in the field of satellite image segmentation (Bhandari et al., 2014). One can efficiently extract features of graphic object by using a Cuckoo search algorithm (Wozniak et al., 2016). Multilevel thresholding of an image can be achieved by optimizing edge magnitude of the image using a Cuckoo search algorithm (Panda et al., 2013a).
This paper presents an efficient design of digital FIR high pass and band stop filters using an adaptive cuckoo search algorithm (ACSA). The important features of ACSA are — (i) the step size is independent and (ii) it is accurately decided from the current fitness value. The step size is decided according to the current fitness value within the iteration process. This increases the convergence speed. The other five global optimizers are also used for optimization. The optimal solutions obtained by the ACSA are compared with the other global optimizers. CEC 2005 benchmark test functions are considered for the comparison. The results are compared in terms of the convergence speed, accuracy, deviation from the desired response, minimum stop-band attenuation and maximum pass-band attenuation. The statistical analysis, i.e. $t$ -Test is performed to claim the superiority of the proposed approach. The simulation results presented in this paper reveal the fact that the performance of the ACSA is better than the other algorithms.
Adaptive neuro-heuristic hybrid model for fruit peel defects detection
2018, Neural Networks
Citation Excerpt :
In Zhang and Zhang (2016) was proposed application of metrics for similarity detection. Heuristics can improve abilities of computer systems for image processing to extract shapes of objects from 2D images, as proposed in Wozniak, Polap, Napoli, and Tramontana (2016). Also visual categorization for shapes in primate cortex is possible if we appropriately model image descriptors as discussed by Sigala and Logothetis (2002).
Fusion of machine learning methods benefits in decision support systems. A composition of approaches gives a possibility to use the most efficient features composed into one solution. In this article we would like to present an approach to the development of adaptive method based on fusion of proposed novel neural architecture and heuristic search into one co-working solution. We propose a developed neural network architecture that adapts to processed input co-working with heuristic method used to precisely detect areas of interest. Input images are first decomposed into segments. This is to make processing easier, since in smaller images (decomposed segments) developed Adaptive Artificial Neural Network (AANN) processes less information what makes numerical calculations more precise. For each segment a descriptor vector is composed to be presented to the proposed AANN architecture. Evaluation is run adaptively, where the developed AANN adapts to inputs and their features by composed architecture. After evaluation, selected segments are forwarded to heuristic search, which detects areas of interest. As a result the system returns the image with pixels located over peel damages. Presented experimental research results on the developed solution are discussed and compared with other commonly used methods to validate the efficacy and the impact of the proposed fusion in the system structure and training process on classification results.
Customizing host IDE for non-programming users of pure embedded DSLs: A case study
2017, Computer Languages, Systems and Structures
Citation Excerpt :
The main advantages of DSLs are high abstraction level, high expressivity due to usage of domain concepts in a language, and involvement of domain experts [2–5]. As a result, applications of the DSL approach can be found in many domains, such as secure logging [6], questionnaires [7], database design [8], feature modeling [9], expert systems [10], etc. Disadvantages are mainly centered around language creation costs and insufficient tool support [2,11].
Pure embedding as an implementation strategy of domain-specific languages (DSLs) benefits from low implementation costs. On the other hand, it introduces undesired syntactic noise that impedes involvement of non-programming domain experts. Due to this, pure embedded DSLs are generally not intended for, nor used by, non-programmers. In this work, we try to challenge this state by experimenting with inexpensive customizations of the host IDE (Integrated Development Environment) to reduce the negative impact of syntactic noise. We present several techniques and recommendations based on standard IDE features (e.g., file templates, code folding, etc.) that aim to reduce syntactic noise and generally improve the user experience with pure embedded DSLs. The techniques are presented using a NetBeans IDE case study. The goal of the proposed techniques is to improve the user experience with pure embedded DSLs with a focus on the involvement of non-programming domain experts (or non-programmers in general). The proposed techniques were evaluated using a controlled experiment. The experiment compared a group using Ruby and non-modified RubyMine IDE versus a group using Java and NetBeans IDE customized to use the proposed techniques. Experiment results indicate that even inexpensive host IDE customizations can significantly alleviate issues caused by the syntactic noise: Java with its inflexible syntax performed better than Ruby with its concise syntax.
Human gait analysis for osteoarthritis prediction: a framework of deep learning and kernel extreme learning machine
2023, Complex and Intelligent Systems
Towards a polynomial approximation of support vector machine accuracy applied to Arabic tweet sentiment analysis
2023, Mathematical Modeling and Computing

View all citing articles on Scopus

View full text

Graphic object feature extraction system based on Cuckoo Search Algorithm

Highlights

Abstract

Introduction

Section snippets

2D Image Key-Points search methods

Novel approach to feature extraction

Feature extraction based on simplified sobel filter and Graphix Cuckoo Search Algorithm

Experimental research results

Final remarks

Expert Systems with Applications

Expert Systems with Applications

Applied Soft Computing

Applied Soft Computing

Information Sciences

Medical Image Analysis

Computer-Aided Design

Expert Systems with Applications

Expert Systems with Applications

Expert Systems with Applications

Expert Systems with Applications