![AN ENHANCED CIRCULAR DETECTION TECHNIQUE - RPSW
USING CIRCULAR HOUGH TRANSFORM
Anisha Ravikumar J
Sree Buddha College of Engineering
Abstract— Earth Surface consist of so many circular features
such as craters, volcanoes and other man-made structures. To
detect these circular features , so many detection techniques
where proposed. All that techniques are based on supervised
learning, unsupervised learning, machine learning etc. Here we
are going to propose a new automatic technique for the detection
of this circular features. Our method is actually based on the
multiplication operation of the original image with the rotated
image. We are combining the RPSW with circular Hough
transform. This method can detect the simple circles and can
also detect the more complex circular features. The method can
also find the area of the circular features.
Keywords- Feature extraction, Image Rotation , Pixel Swapping
Object detection, Remote sensing.
1. INTRODUCTION
The circular features, like impact craters or volcanoes, or
geological domes as well as the man-made
structures/symbols, are found on the satellite images. This
circular features are studied to get the relative age of the
planetary surface. Manually finding the circular features are
very difficult and it need more human skill. So in-order to
avoid that problem, so many automatic detection algorithms
are there. Image based and tomography based detection are
the two main approaches in the field of circular feature
detection.
Hough Transform and Wavelet transforms are the commonly
used techniques used for the image based detection
algorithms. Watershed and Terrin Derivatives are using in
topology based detection. Unsupervised detection of circular
features on the images depend on pattern recognition.
The main issues in lunar research is to find the age of lunar
surface by estimating the density of craters per unit area.so it
should detect the circular features clearly and also the shape
and size of the circular features should found clearly.
Detecting this type of circular features are very important for
finding the age of moon surface or other planetary surface.
So many automatic feature detection techniques have been
developed on the basis of pattern-matching tor Hough
Transform [1]–[2]. But in all the existing methods so many
problems are there. Main problem is about the accuracy of
the algorithm as well as the need of additional pre-processing
before applying the detection algorithms.
In this paper, we propose a new automatic method for the
detection and extraction of circular features from the satellite
images. This method is an image based detection technique.
We are doing this in black-and-white images (b binary image)
. This method is based partly on the pixel swapping (PSW)
for pattern extraction. [3].Here we are combining the
Rotational pixel swapping algorithm with the Hough
Transform. We are performing rotation of images as well as
the multiplication of images. The method can detect the
simple circular features as well as the more complex circular
features.
II. RPSW WITH HOUGH TRANSFORM FOR
DETECTION OF CIRCULAR FEATURES IN
SATELLITE IMAGES
With the combined RPSW and Hough Transform we can
accurately detect the circular features from the planetary
images.
A. RPSW
It means Rotational pixel swapping. The PSW algorithm
will extract the patterns by an inter-picture operation between
the original image with the translated image. In PSW
algorithm, as a rule of the translation operation target pattern
ISBN:978-1535061506
www.iaetsd.in
Proceedings of ICRMET-2016
©IAETSD 201621](https://image.slidesharecdn.com/iaetsdanenhancedcirculardetectiontechnique-rpswusingcircularhoughtransform-160706025038/85/Iaetsd-an-enhanced-circular-detection-technique-rpsw-using-circular-hough-transform-1-320.jpg)
![Fig: 2.5: radius of the circle is detected and all other shapes are
eliminated.(After applying Circular Hough Transform)
When we apply this Circular Hough Transform in the
resultant RPSW output, we can find the correct radius as
well as the area of the particular circle. Using RPSW we can
find the centre of the circle and then can find the radius by
apply Circular Hough Transform.
III. CONCLUSION
In this paper, we are introducing a new method for finding
the centre, radius and area of the circular features in the
satellite images. Here both the RPSW Algorithm and the
Circular Hough Transform is combined to get the result
Acknowledgment
I thank almighty for giving me the strength and courage to
take a good area and to do research in that area. I am thankful
to my project guide for his suggestion sand help me by giving
all the useful facilities like Internet access as well as books,
which is more important to me. I am also very thankful to all
staffs in the Department of Computer Science & Engineering
of Sree Buddha College of Engineering, pattoor, Alappuzha
References
[1] P. V. C. Hough, “Method and means for recognition complex
patterns,” U.S. Patent 3 069 654, Dec. 18, 1962.
[2] P. Pina, J. Saraiva, and T. Barata, “Automatic recognition of
Aeolia ripples on Mars,” presented at the Lunar Planetary
Science Con e The Woodlands, TX, USA, Mar. 2004, Paper
1621J.
[3] J. Iisaka and T.Sakurai-Amano, “An application of pixel
swapping technique to remote sensing,” presented at the
Proceeding Asian ConfConference Remote Sensing, Singapore,
2000, Paper OMPOO-13
[4] B. Ramachandran, C. O. Justice, and M. J. Abrams, Land Remote
Sensing and Global Environmental Change. NASA’s Earth
Observing System and the Science of ASTER and MODIS. New
York, NY, USA: Springer-
[5] H. J. Melosh, Impact Cratering— A Geologic Process. New
York, NY,
[6] M. Abrams, “The Advanced Spaceborne Thermal Emission and
Reflection .Radiometer (ASTER): Data products for the high
spatial resolution imager on NASA’s Terra platform,” Int. J.
Remote Sens., vol. 21, no. 5, pp. 847–859, Nov. 2000.
[7] P. Lambert, J. F. McHone, Jr., R. S. Dietz, M. Briedj, and M.
Djender, “Impact and impact-like structures in Algeria. PartII
Multi-ringed structures,”Meteoritics, vol. 16, no. 3, pp. 203–227,
Sep. 1981
ISBN:978-1535061506
www.iaetsd.in
Proceedings of ICRMET-2016
©IAETSD 201624](https://image.slidesharecdn.com/iaetsdanenhancedcirculardetectiontechnique-rpswusingcircularhoughtransform-160706025038/85/Iaetsd-an-enhanced-circular-detection-technique-rpsw-using-circular-hough-transform-4-320.jpg)
Iaetsd an enhanced circular detection technique rpsw using circular hough transform
- 1. AN ENHANCED CIRCULAR DETECTION TECHNIQUE - RPSW USING CIRCULAR HOUGH TRANSFORM Anisha Ravikumar J Sree Buddha College of Engineering Abstract— Earth Surface consist of so many circular features such as craters, volcanoes and other man-made structures. To detect these circular features , so many detection techniques where proposed. All that techniques are based on supervised learning, unsupervised learning, machine learning etc. Here we are going to propose a new automatic technique for the detection of this circular features. Our method is actually based on the multiplication operation of the original image with the rotated image. We are combining the RPSW with circular Hough transform. This method can detect the simple circles and can also detect the more complex circular features. The method can also find the area of the circular features. Keywords- Feature extraction, Image Rotation , Pixel Swapping Object detection, Remote sensing. 1. INTRODUCTION The circular features, like impact craters or volcanoes, or geological domes as well as the man-made structures/symbols, are found on the satellite images. This circular features are studied to get the relative age of the planetary surface. Manually finding the circular features are very difficult and it need more human skill. So in-order to avoid that problem, so many automatic detection algorithms are there. Image based and tomography based detection are the two main approaches in the field of circular feature detection. Hough Transform and Wavelet transforms are the commonly used techniques used for the image based detection algorithms. Watershed and Terrin Derivatives are using in topology based detection. Unsupervised detection of circular features on the images depend on pattern recognition. The main issues in lunar research is to find the age of lunar surface by estimating the density of craters per unit area.so it should detect the circular features clearly and also the shape and size of the circular features should found clearly. Detecting this type of circular features are very important for finding the age of moon surface or other planetary surface. So many automatic feature detection techniques have been developed on the basis of pattern-matching tor Hough Transform [1]–[2]. But in all the existing methods so many problems are there. Main problem is about the accuracy of the algorithm as well as the need of additional pre-processing before applying the detection algorithms. In this paper, we propose a new automatic method for the detection and extraction of circular features from the satellite images. This method is an image based detection technique. We are doing this in black-and-white images (b binary image) . This method is based partly on the pixel swapping (PSW) for pattern extraction. [3].Here we are combining the Rotational pixel swapping algorithm with the Hough Transform. We are performing rotation of images as well as the multiplication of images. The method can detect the simple circular features as well as the more complex circular features. II. RPSW WITH HOUGH TRANSFORM FOR DETECTION OF CIRCULAR FEATURES IN SATELLITE IMAGES With the combined RPSW and Hough Transform we can accurately detect the circular features from the planetary images. A. RPSW It means Rotational pixel swapping. The PSW algorithm will extract the patterns by an inter-picture operation between the original image with the translated image. In PSW algorithm, as a rule of the translation operation target pattern ISBN:978-1535061506 www.iaetsd.in Proceedings of ICRMET-2016 ©IAETSD 201621
- 2. with in the shift table is used. In this method the pattern of a similar size in the shift table can be extracted. So, while using this PSW, we should prepare patterns with all the possible sizes for the shift table. To avoid that problem Rotational Pixel Swapping (RPSW) algorithm has come. RPSW extracts rotational symmetric pattern by multiplying the original image with the rotated images. Let A(x, y) be the original image. We can apply this RPSW algorithm only in binary images, so we want to covert the image A(x, y) in to a binary image. If the pixel value A(x, y)= 0 then it is a black pixel. If the pixel value A(x, y)=1,thenit will be a white image. x, y are the horizontal axes and the vertical axes. (0,0) is the top left point of the image. We will produce so many images from the original image A(x, y) by rotating it upon different angles. The resultant rotated image is represented as Bxn,yn,φ(x, y). Here φ(x, y) is the rotation angle. After finding the rotated images, Mφ(x, y) will find for each φ for producing he extracted image and let it be C. Mφ(x, y) = A(x, y) · Bxn,yn,φ(x, y) (1) After this operation we can see an enhancement in the circular , all the other shapes are eliminated. Now we need to find the centre of the circle .Below equation is used for that. R(x, y) = ∑ ∑ Bx,y,φ = k·Δφ(i, j) (2) where (x, y) is the rotation centre for the calculation of the rotated image Bx,y,φ, Δφ is an incremental angle of rotation, and N is the total number of rotation images. If the rotation centre (x, y) is near the CRSP, R is large. If the rotation centre is far away from the CRSP, R is small. Thus, a relatively high R value at a point indicates that the point is the centre of a circular feature. We are calculating R(x, y) for following region (i, j) only: lmin < √(i − x)2 + (j − y)2 < lmax (3) lmax and lmin are the maximum local RPSW radii and minimum local RPSW radii By using lmax and lmin we can reduce the processing time T for (x,y). Fig 2.1: A colour composite image. When we implement the calculation of R practically, N rotation images should generate first by Δφ around the center of the image and shifted them to the location needed for different centres of the rotation. All the selected points (xn, yn) whose R values are bigger than threshold value f · Rmax, here f is threshold fraction and Rmax is maximum value in R. Then we calculate Cn(x, y) for (xn, yn) for each Rotational Symmetric Pattern whose R value is nth highest in R .It is done by using following equation. N Cn(x, y) =∑ A(x, y) · Bxn,yn,φ = k·Δφ(x, y). k=1 Here also we are calculating Cn(x, y) for the region (i, j) only defined by the equation (3) and after that summed up all Cn(x, y) for the Rotational Symmetric Pattern s which having R > f · Rmax, i.e., C(x, y) = ∑n Cn(x, y). In RPSW method some problems are there that we identified. We cannot find the area of the circular feature, we can find only the centre of the circular features. To solve this problem, we proposed a circular feature detection algorithm. In this algorithm, it should produce the rotational images, and then it should perform the multiplication and then perform summation. After doing this, it should apply the Circular Hough transform on the resultant image. Here, we are the combining the RPSW algorithm with the Hough Transform. ISBN:978-1535061506 www.iaetsd.in Proceedings of ICRMET-2016 ©IAETSD 201622
- 3. Fig 2.3: : RPSW combined with Hough Transform Circular Hough Transform is used to find the circular objects in the noisy images. The method is based on the conversion of images to binary images. Edge detection techniques such as sobel operator or canny operators are using in the Circular Hough Transform. A voting procedure is used in Circular Hough Transform. Circular Hough Transform is based on the equation of circles r² = (x – a)² + (y – b)² (4) a and b centre coordinates and r is the radius of the circle. x = a + r*cos(θ) (5) y = b + r*sin(θ) (6) In CHT radius is set to a constant or we are providing a particular rage of values. So that we can reduce the complexity of using the algorithm. A circle is drawn at each edge point with that edge point as the origin point and radius is r. An array(3D) is used with first 2Ds represents the coordinates of circle and last third represents the radius. Whenever a circle is drawn with the desired radii over each of the edge point, values in accumulator array will increased. Accumulator, keeps the count of how many circles are passing through the coordinates of each of the edge points, and it will produce a vote to find highest count. The coordinates of the centre of the circles in the images are the coordinates with the highest count. Fig 2.3 : Sample image to do the process Fig 2.4: Centre detected only in the circular region (output of RPSW) Fig 2.2: Block Diagram For The RPSW ISBN:978-1535061506 www.iaetsd.in Proceedings of ICRMET-2016 ©IAETSD 201623
- 4. Fig: 2.5: radius of the circle is detected and all other shapes are eliminated.(After applying Circular Hough Transform) When we apply this Circular Hough Transform in the resultant RPSW output, we can find the correct radius as well as the area of the particular circle. Using RPSW we can find the centre of the circle and then can find the radius by apply Circular Hough Transform. III. CONCLUSION In this paper, we are introducing a new method for finding the centre, radius and area of the circular features in the satellite images. Here both the RPSW Algorithm and the Circular Hough Transform is combined to get the result Acknowledgment I thank almighty for giving me the strength and courage to take a good area and to do research in that area. I am thankful to my project guide for his suggestion sand help me by giving all the useful facilities like Internet access as well as books, which is more important to me. I am also very thankful to all staffs in the Department of Computer Science & Engineering of Sree Buddha College of Engineering, pattoor, Alappuzha References [1] P. V. C. Hough, “Method and means for recognition complex patterns,” U.S. Patent 3 069 654, Dec. 18, 1962. [2] P. Pina, J. Saraiva, and T. Barata, “Automatic recognition of Aeolia ripples on Mars,” presented at the Lunar Planetary Science Con e The Woodlands, TX, USA, Mar. 2004, Paper 1621J. [3] J. Iisaka and T.Sakurai-Amano, “An application of pixel swapping technique to remote sensing,” presented at the Proceeding Asian ConfConference Remote Sensing, Singapore, 2000, Paper OMPOO-13 [4] B. Ramachandran, C. O. Justice, and M. J. Abrams, Land Remote Sensing and Global Environmental Change. NASA’s Earth Observing System and the Science of ASTER and MODIS. New York, NY, USA: Springer- [5] H. J. Melosh, Impact Cratering— A Geologic Process. New York, NY, [6] M. Abrams, “The Advanced Spaceborne Thermal Emission and Reflection .Radiometer (ASTER): Data products for the high spatial resolution imager on NASA’s Terra platform,” Int. J. Remote Sens., vol. 21, no. 5, pp. 847–859, Nov. 2000. [7] P. Lambert, J. F. McHone, Jr., R. S. Dietz, M. Briedj, and M. Djender, “Impact and impact-like structures in Algeria. PartII Multi-ringed structures,”Meteoritics, vol. 16, no. 3, pp. 203–227, Sep. 1981 ISBN:978-1535061506 www.iaetsd.in Proceedings of ICRMET-2016 ©IAETSD 201624