IMPLEMENTATION OF IHS FUSION TECHNIQUE AND COMPARATIVE ANALYSIS WITH PCA FUSION TECHNIQUE FOR COTTON CONTAMINANTS DETECTION

1VIJENDRA KUMAR, 2MAITERYEE DUTTA, 3SURENDER SINGH SAINI

1M.E. Student NITTTR Chandigarh, 2Professor and Head CSE Dept. NITTTR Chandigarh, 3Senior Scientist, CSIO Chandigarh

Abstract- Contaminants in cotton have serious effect on the quality of cotton fiber. The removal of cotton contaminants by manual system requires a lot of manpower and time consuming proces therefore, in textile industries, automatic cotton contamination detection is used. Digital image processing algorithm based on machine vision provides efficient and accurate detection of contaminants. Various techniques in this field developed and implemented based on Co-occurrence Matrix Contrast Information, on the Basis of Wavelet, Neighborhood Gradient Based on YCbCr , using Intensity And Hue Properties, Based on RGB Space Model, using YDbDr Color Space, using X-Ray microtomographic image analysis, Cotton Using Near Infrared Optimal Wavelength Imaging, PCA for Detecting contaminants in Cotton and comparision with various color spaces. Intensity Hue Saturation (IHS) fusion technique has not been implemented for detecting different types of contaminants. Using PCA and IHS fusion technique, we developed new methods for detecting contaminants’ in cotton efficiently and effectively. In this research we concluded that the IHS fusion applied on HSV is gives best result in terms of all contaminants detection, less number of false target detection and the clarity in output image.

Keywords- Cotton contaminants; Detection; PCA Fusion; IHS Fusion; YCbCr; YDbDR; IHS; HSV; Comparison

I. INTRODUCTION

Cotton contaminants refer to foreign fibers in cotton such as animal fiber, polypropylene twines, field plastic film, plant leaves, human hairs etc. So these contaminants in cotton can badly influence the level of quality for ginned cotton and its commodity market price, they affects the cotton production process and fiber dyeing evenness.

There are lots of scholars from home and abroad study Figure 1: Cotton contaminants detection system on how to purge away contaminants from cotton efficiently. In this paper, we separate gray and color II. RELATED WORK information in YCbCr, YDbDr, IHS, and HSV color space, and design a cotton contaminants detection A fairly large number of cotton fibers recognition algorithm using PCA fusion and IHS fusion researches are based on RGB color space. Reference techniques. Cotton contaminants detection system is proposed a physical model arithmetic color image shown in fig. 1, which consists of three parts. clustering algorithm in RGB color space.

First part is cotton image collection part,e second part Reference proposed a method which uses is the contaminants detection part and third part is the decomposition of 2-D image into multi-layer wavelet reporting port by which machine instruct the high by using wavelet packet 2-D, and selects the best base pressure air nozzles to saperate contaminants from structure to set entropy type and value for compression cotton layer. and de-noising, then, analyzes and processes the wavelet coefficient to carry on multi-scale In first part we collect images of raw cotton layers on reconstruction to the Information numbers obtained the time of ginning and images are converted into from them, finally, adopts wiener smoothing filtering desired color space and after the needed processing of and binary processing technology. the images of cotton the resulted image will be produced which will collect the information about Conversion of images from RGB to YCbCr color contaminants or non contaminants. Finally after space, obtained gray image and color component processing images will be printed on printer and air images. For gray image an image sharpening nozzles to separate contaminants. algorithm based on neighborhood gradient is applied

Proceedings of 7th IRF International Conference, 12th October 2014, Goa, India, ISBN: 978-93-84209-57-5 82 Implementation of IHS Fusion Technique And Comparative Analysis With PCA Fusion Technique For Cotton Contaminants Detection the iterative thresholding is used for segmentation and the direction of the maximum variance. The second simple fusion technique is appliec to get fused image. principal component is forced to lie in the subspace vertical (perpendicular) of the first. Within this In other paper for detection of cotton contaminants, subspace, this component points the direction of the novelty of the proposed approach lies in selection maximum variance. The third principal component is the channel background in the cotton model space to taken in the maximum variance direction in the solve the interference of background for the detection. subspace vertical to the first two and so on. The image fusion using Principal component analysis is Next the implementation and comparative analysis of represented in Figure 2. the YDbDr, YPbPr, YCbCr and HSI color spaces for the detection of contaminants from the cotton is the Reference proposed a method in which input images main objective of this paper to detect small (images to be fused) A (x, y) and B (x, y) are arranged contaminants or insects from the cotton with more in two column vectors and their empirical means are clarity which was not possible in YCbCr and HSI color subtracted. The resulting vector has a dimension of n x space. 2, where n is length of the each image vector. Compute the Eigenvector and Eigen values for this resulting And finally The paper presents the implementation vector are computed and the eigenvectors and comparative analysis of the YDbDr, YCbCr and corresponding to the larger eigenvalue obtained. The HSI color spaces for the detection of contaminants normalized components P1 and P2 (i.e., P1 + P2 = 1) from the cotton. One of the main objectives of this are computed from the obtained eigenvector and the paper is to detect small contaminants or insects from final fused image is: I (fused image) = AP1+BP2 the cotton with more clarity which was not much possible in YCbCr and HSI color space and YDbDr A. PROPOSED PCA ALGORITHM with normal fusion method. Results shows the 1. Get image from source and transform into desired comparison between these three color spaces on the color space (YCbCR, YDbDr or IHS color space). basis of , mean, variance, and standard deviation, 2. Extract layers (three layers ) from transformed Which shows that YDbDr is better than other three image (Y, CB, CR OR Y, DB, DR OR I, H, S). color spaces and more accurately detected using PCA 3. Binarization of layers ( Y,Cb,Cr or Y, Db,Dr or I,H, fusion technique, But in this paper false targets S ). appears vastly. 4. Repeat steps 5 to 6 to fuse layers 5. F1= PCA(Layer-1 and Layer-2) In this paper, we separate gray and color information 6. F2=PCA(F1 and Layer-3) in YCbCr, YDbDr, IHS, and HSV color space, and and get the final fused image (F2). design a cotton contaminants detection algorithm 7. Calculation of desired parameters. using PCA fusion and IHS fusion techniques to detect all contaminants with less number of false targets.

III. PRINCIPAL COMPONENT ANALYSIS FUSION

Fig 3. Proposed PCA fusion diagram.

B. IHS FUSION FOR COTTON CONTAMINANTS DETECTION Fig 2. PCA fusion. IHS transform is a frequent used fusion method,I refers to intensity; H refers to hue and S refers to In PCA fusion the principal components are used to saturation. As human eyes can get higher resolution fuse images. first principal component is taken along on image intensity than on hue and saturation, so

Proceedings of 7th IRF International Conference, 12th October 2014, Goa, India, ISBN: 978-93-84209-57-5 83 Implementation of IHS Fusion Technique And Comparative Analysis With PCA Fusion Technique For Cotton Contaminants Detection images of three spectral bands with low spatial IV. EXPERIMENTAL RESULTS resolution in RGB description system can be converted to IHS color system, after replacing the I In this experiment, operating environment’s CPU is component with high spatial resolution image INTEL® core™ i3-3217U, and internal memory is 4 recorded as I', then apply I'HS to the RGB inverse GB, Operating system is Windows 7, Software transform. In this way, the result image can have the environment is Matlab 7.12.0.635 (R2011a). The characteristic of spatial resolution level of training samples are the cotton samples' original high-resolution images with contaminants. Simulation results on fused images, visual observation, comparison tables of parameters and charts are given below.

Fig 4. IHS fusion diagram Fig. 6: Original image 1 C. PROPOSED IHS ALGORITHM 1. Get image from source and transform into desired color space (YCbCR, YDbDr,IHS or HSV color space). 2. Extract layers from transformed image (Y, Cb, Cr or Y, Db, Dr, I, H, S or H, S, V). 3. Binarization of layers ( Y, Cb, Cr or Y, Db, Dr, I, H, S or H, S, V ). 4. Transform original RGB image to Grayscale 5. The histogram equalization of Gray scale image and Intensity image of transformed image. 6. Replace the Intensity image with new Intensity image 7. Inverse transformation from YCbCr or YDbDr or IHS or HSV to RGB image 8. Transform RGB image to Grayscale image. Fig. 7: PCA fused images of image 1

9. Binarization of Gray scale image and get final image 10. Calculation of desired parameters.

Fig. 8: IHS fused images of image 1

Fig 5. Proposed IHS fusion diagram. Fig. 9: Original Image 2

Proceedings of 7th IRF International Conference, 12th October 2014, Goa, India, ISBN: 978-93-84209-57-5 84 Implementation of IHS Fusion Technique And Comparative Analysis With PCA Fusion Technique For Cotton Contaminants Detection Table3. Visual Observation in PCA fusion for fused image 2 Parameters YCbCr YDbDr IHS HSV Detect all Partiall Partiall Contaminant No Yes y y s False Yes No Yes Yes Targets Light Color Partiall Partiall Contaminant No Yes y y Detection Visual Clarity of No No No Yes contaminants FIg. 10: PCA fused images of image 2 Table4. Visual Observation in IHS fusion for fused image 2 Parameters YCbCr YDbDr IHS HSV

Detect all Partiall Partiall Contaminant No Yes y y s False No More More Less Targets Light Color Partiall Partiall Contaminant No Yes y y Detection Visual Clarity of No No No Yes contaminants FIg. 11: IHS fused images of image 2

Table1. Visual Observation in PCA fusion for fused image 1 Parameters YCbCr YDbDr IHS HSV

Detect all Contaminants No Yes Partially Partially False Targets Yes No Yes No Light Color Contaminant No Partially Partially Partially Detection Visual Clarity of contaminants No No Yes Yes

Table2. Visual Observation in IHS fusion for fused image 1 Parameters YCbCr YDbDr IHS HSV Fig. 12: Execution time for image 1 (PCA & IHS) Detect all Contaminants No No No Yes Table5. Execution time(sec.) taken by IHS fusion for image-1 False Less Less No No Fusion YCbCr YDbDr IHS HSV Targets number number Technique color color color color Light Color space space space space Contaminant No No No Yes PCA 0.08461 0.05904 0.11007 0.15172 Detection Fusion 9 8 3 9 Visual Clarity IHS 0.12454 0.07612 0.25440 0.23811 of No No No Yes contaminants Fusion 1 9 1 3

Proceedings of 7th IRF International Conference, 12th October 2014, Goa, India, ISBN: 978-93-84209-57-5 85 Implementation of IHS Fusion Technique And Comparative Analysis With PCA Fusion Technique For Cotton Contaminants Detection Visual observation shows the comparison between these four color spaces used in PCA and IHS fusion techniques. we concluded that all the contaminants in cotton are detected in IHS fusion with HSV color space with less number of false targets and clarity of contaminants. Furthermore cotton contaminants detection algorithm still need to be optimize for execution time and develop contaminants detection technique using other fusion techniques like wavelet, HPF etc.

REFERENCES

[1] Xianying Feng, Chengliang Zhang, Bingsheng Yang and Lei

Fig. 13: Execution Time for image 2 (PCA & IHS) Li, “Foreign fiber recognition and detection algorithm based on RGB color space,” Journal of Shandong University (Engg Science Table6. Execution time(sec.) taken by IHS fusion for [2] Chengliang Zhang, Xianying Feng, Lei Li, Yaqing Song image-2 ,“Detection of Foreign Fibers in Cotton on the Basis of Fusion YCbCr YDbDr IHS HSV Wavelet”, 2nd International Conference on Signal Processing Technique color color color color Systems (ICSPS), Dalian Page(s): V2-304 - V2-308 2010 space space space space IEEE. PCA 0.11310 0.084831 0.054928 0.152198 Fusion 2 [3] Zhang, Xianying Feng, Lei Li, Yang Song, “Identification of Cotton Contaminants Using Neighborhood Gradient Based on IHS 0.24968 YCbCr Color Space”, 2010 2nd international conference on 0.122040 0.086507 0.242608 Fusion 2 signal processing systems, Dalian Page(s): V3-733 - V3-738 IEEE. For BAR GRAPH: [4] Tingting Xie Yalin Gu Tao Sha Yapeng He,” A Method for Detection of Foreign Body in Cotton Based on RGB Space *Bar 1 and 2 for YCbCr color space with PCA and Model”, 2nd International Conference on Artificial IHS fusion Intelligence, Management Science and Electronic Commerce *Bar 3 and 4 for YDbDr color space with PCA and (AIMSEC), Deng Leng Page(s): 31 - 33 2011 IEEE IHS fusion [5] Gurjinder Singh Sahdra, Kamaljot Singh Kailey, ” Detection *Bar 5 bar 6 for IHS color space with PCA and IHS of Contaminants in Cotton by using YDbDr Color Space” fusion Int.J.Computer Technology & Applications,May-June 2012, Vol 3 (3), 1118-1124 *Bar 7 and 8 for HSV color space with PCA and IHS fusion [6] Damandeep kaur, Mr.SunilKhullar,” Principal Component Analysis for Detecting Foreign Fiber Contaminant in Cotton with Comparative Analysis”, ISSN 2319-7080, International CONCLUSION AND FUTURE SCOPE Journal of Computer Science and Communication Engineering Volume 3 issue 1. The paper presents the implementation and [7] Prashant Sangulagi, Raj reddy, “Comparative Analysis of comparative analysis of PCA fusion and IHS Fusion Different Image Fusion Techniques for Sensor Application”, Indian Journal of Engineering, Volume 8, Number 19, on YCbCR, YDbDr, IHS and HSV color spaces for the February 19, 2014 pp, 23-30, detection of contaminants from the cotton. One of the [8] Wang Guang-jun, Xiao Qiu-ping, Fu Mei-chen, Wang zeng, main objective of this paper is to detect all “Data Fusion for Multi-sources Remotely Sensed Images contaminants from the cotton with more clarity and Based on IHS Transformation and Wavelet Analysis”, 2009 less number of false targets. comparison tables of International Conference on Environmental Science and Information Application Technology.



Proceedings of 7th IRF International Conference, 12th October 2014, Goa, India, ISBN: 978-93-84209-57-5 86