Table of contents
2. Least squares edge fitting extraction method based on region of interest
2.1 Extraction of regions of interest
2.1.2 Interactive extraction of regions of interest
2.2 Edge extraction of circular signs
2.3 Least squares edge fitting and circular mark positioning
3. Extraction method based on image grayscale and geometric properties
3.1 Extraction of circular marks
3.2 Image positioning based on geometric properties
Summary
Aiming at the problem of extracting and positioning circular marks on complex background images in the calibration of visual measurement cameras at the shooting range, two different circular mark extraction methods were studied. One is an interactive extraction method based on the area of interest, and the other is Extraction method based on image grayscale and geometric morphology.
The extraction results of the circular mark show that the first method can smoothly extract the edge of the circular mark, but it is greatly affected by the imaging situation of the circular mark. The edge extraction effect of the second method is affected by the imaging situation of the circular mark. Small, but the extracted edges are zigzag! Some pixels of the circular mark are misjudged as background, and there is a certain error with the actual imaging of the circular mark.
I. Introduction
The circular logo has rotation invariance and will be imaged as an ellipse after being photographed at a certain angle! It is more robust than linear or point-shaped logos.
Due to the shooting angle, the circular mark appears as an ellipse on the image, so the extraction of the circular mark is mainly an ellipse detection problem. Hough transform and its improved algorithm are commonly used ellipse detection methods. When detecting ellipses, methods based on Hough transform need to map the image space to parameter space for multi-dimensional statistical calculations, which requires a large amount of calculation and low efficiency. At the same time, when performing on-site visual measurements, the image background is relatively complex, and in static explosion experiments it will be interfered by shock waves, dust and other complex factors, so the application of Hough transform is more difficult.
In response to these problems, the interactive extraction method based on the region of interest (ROI) and the extraction method based on image grayscale and geometric morphology were studied to extract and locate the circular sign image respectively. These two methods have a small amount of calculation and can quickly complete the positioning of the circular mark when the number of mark points is known.
2. Least squares edge fitting extraction method based on region of interest
2.1 Extraction of regions of interest
2.1.1 Region of Interest
The area of interest refers to the area in images and videos that can attract the attention of the human visual system! These areas often have significant brightness, color, shape, texture and other characteristics compared to the surrounding environment. The area of interest contains the main information and key information of the image, so the processing of the area of interest can approximately replace the processing of the original image, which can greatly reduce the amount of calculation and reduce the complexity of image processing.
The methods for extracting regions of interest mainly include human-machine interactive methods, fixation point-based methods, visual attention model-based methods, and specific object segmentation-based methods, etc. Among them, the human-computer interactive method is simple to operate and easy to implement, so the interactive method is used to extract the area of interest where the calibration circle is located.
2.1.2 Interactive extraction of regions of interest
Using MATLAB software as an image processing tool, the waitforbuttonpress function is used to obtain the coordinates of the point where the mouse clicks on the image to obtain the selected area.
2.2 Edge extraction of circular signs
The steps for ellipse detection are as follows:
1) Enter the area of interest of the circular markI1 and record Position coordinates of the upper left corner of the area of interest(x0, y0);
2) Filtering, perform median filtering on I1 to reduce the impact of noise on the ellipse boundary without losing target information. , enhance the connectivity between ellipse edge pixels;
3) Binarization processing, binarize the image, set the pixel value of the background area to 0, and set the pixel value of the circular mark area to 1;
4) Area filling, filling the inside of the closed area and eliminating holes on the edges;
5) Interference area elimination, detect the connected areas on the image and count the number of pixels in each connected area as the characteristic area of the connected area. The area with the largest area is the area where the circular sign is located, and other smaller areas are eliminated;
6) Edge extraction, the Canny operator has the characteristics of high precision and fast operation speed, so the Canny operator is used to extract the edge of the circular sign
;
7) According to the edge coordinates and the position coordinates of the upper left corner extracted in step 6) (x0, y0), the original The coordinates of the edge of the circular mark in the image on the image.
2.3 Least squares edge fitting and circular mark positioning
After extracting the edge of the circular mark, the position coordinates of all edge points can be obtained, and the least squares fitting method is used to restore the shape of the circular mark
. The general algebraic form of the elliptic curve equation can be expressed as
Rewritten into matrix form, that is
3. Extraction method based on image grayscale and geometric properties
3.1 Extraction of circular marks
The image is preprocessed according to the grayscale characteristics of the image, then edge detection is performed, and the closed contour is filled to obtain multiple areas containing the circular mark, and then the interference areas are eliminated according to the area characteristics of the circular mark. The specific processing process is as follows:
1) Median filtering to eliminate the influence of noise and make the edges clearer;
2) Transform the grayscale of the image to enhance the contrast between light and dark parts of the image;
3) Perform edge detection on the image to obtain multiple contours containing circular marks;
4) Hole area filling, fill the area with closed outline, and leave the area with unclosed outline as it is;
5) Eliminate linear areas on the image according to the circularity of the area. The circularityC reflects the degree to which an ellipse is close to a circle. , the closer the value is to 1, the closer the shape of the region is to a circle. The circularity is calculated based on 
, S is the area of the region, and L is the perimeter of the region. . Set the upper and lower thresholds C1 and C2 to eliminate the areas where C>C1 and C<C2;
6) Perform connected area detection on the processed image, and mark the area where the circular mark is located, so that the location area of each circular mark can be segmented.
3.2 Image positioning based on geometric properties
According to the area of the circular mark image extracted in 3.1, the circular mark is positioned on the image based on the geometric properties of the circular mark area. The position of the circular mark on the image is represented by the centroid of the circular mark area, and the grayscale weighted centroid method is used to solve the position coordinates of its center. The formula is as follows:
Summarize
This paper studies two circular landmark extraction methods, namely the least squares edge fitting method based on the region of interest and the method based on image grayscale and geometric properties, and compares the edge extraction effects and positioning results of the two methods.
The least squares edge fitting method based on the region of interest can effectively reduce the amount of image processing calculations by extracting the region of interest. When the circular sign imaging is relatively regular, it can obtain a smooth elliptical edge that is consistent with the actual imaging of the circular sign. However, this method Manual assistance is required, and when the circular mark is irregularly imaged, the edge fitting effect is poor.
The method based on image grayscale and geometric properties is based on the grayscale features and geometric features of the circular sign. It can automatically draw the outline of the circular sign according to the changes in image grayscale without manual intervention, and the edge extraction effect is not good. Affected by the imaging situation of the circular sign, the edges extracted by this method are zigzag outlines, and some pixels on the circular sign are misjudged as background, which has a certain error with the actual imaging situation of the circular sign.