HALCON-从入门到入门-霍夫识别直线

1.废话

霍夫变换是一种特征检测，被广泛应用在图像分析、计算机视觉以及数位影像处理。霍夫变换是用来辨别找出物件中的特征，例如：线条。他的算法流程大致如下，给定一个物件、要辨别的形状的种类，算法会在参数空间中执行投票来决定物体的形状，而这是由累加空间里的局部最大值来决定。

　　现在广泛使用的霍夫变换是由RichardDuda和PeterHart在公元1972年发明，并称之为广义霍夫变换，广义霍夫变换和更早前1962年的PaulHough的专利有关。经典的霍夫变换是侦测图片中的直线，之后，霍夫变换不仅能识别直线，也能够识别任何形状，常见的有圆形、椭圆形。1981年，因为DanaH.Ballard的一篇期刊论文“Generalizing the Hough transform to detect arbitrary shapes”，让霍夫变换开始流行于计算机视觉界。

2.实现效果

3.代码解析

1.读取图片

read_image (Image, 'fabrik')

2.裁剪出白色部分

rectangle1_domain (Image, ImageReduced, 170, 280, 310, 360)

3.执行索贝尔滤波

sobel_dir (ImageReduced, EdgeAmplitude, EdgeDirection, 'sum_abs', 3)

4.二值化出区域

threshold (EdgeAmplitude, Region, 55, 255)

5.裁剪出二值化的区域内的像素

reduce_domain (EdgeDirection, Region, EdgeDirectionReduced)

6.执行霍夫直线检测

hough_lines_dir (EdgeDirectionReduced, HoughImage, Lines, 4, 2, 'mean', 3, 25, 5, 5, 'true', Angle, Dist)

7.生成直线

gen_region_hline (LinesHNF, Angle, Dist)

4.文件

* Detect lines in an image with the help of the Hough transform
* using the edge direction as additional information and return it both
* in HNF and as regions
* 
read_image (Image, 'fabrik')
rectangle1_domain (Image, ImageReduced, 170, 280, 310, 360)
* Detect edges (amplitude and direction) using the Sobel operator
sobel_dir (ImageReduced, EdgeAmplitude, EdgeDirection, 'sum_abs', 3)
dev_set_color ('red')
threshold (EdgeAmplitude, Region, 55, 255)
* Reduce the direction image to the edge region
reduce_domain (EdgeDirection, Region, EdgeDirectionReduced)
* Start the Hough transform using the edge direction information
hough_lines_dir (EdgeDirectionReduced, HoughImage, Lines, 4, 2, 'mean', 3, 25, 5, 5, 'true', Angle, Dist)
* Store input lines described in HNF
gen_region_hline (LinesHNF, Angle, Dist)
dev_display (Image)
dev_set_colored (12)
* Display the lines
dev_set_draw ('margin')
dev_display (LinesHNF)
* Display the edge pixels that contributed to the corresponding lines
dev_set_draw ('fill')
dev_display (Lines)