dyn_threshold(动态阈值)解析

在图像处理的时候会遇到背景和ROI区域难以提取的情况。不能简单的使用一个全局阈值来对图像进行分割。此时会用到动态阈值分割的方法:

转自:dyn_shreshold(动态阈值)解析

1. 算子结构

首先看dyn_threshold 算子参数 
dyn_threshold(OrigImage, ThresholdImage : RegionDynThresh : Offset, LightDark ) 
参数分析: 
OrigImage:需要进行阈值分割的原始图像

ThresholdImage: 通过一定图像预处理得到的阈值图像(这个预处理通常是对原图应用mean_image/gauss_image/binomial_filter 处理后的图像)

RegionDynThresh: 对原图进行阈值分割后输出图像

Offset: 是一个阈值调节值,在做对比时,需要和ThresholdImage上某一点像素值同时作用。

LightDark:是一个关键选择,’dark’, ‘light’, ‘equal’, ‘not_equal’一共有4种选择,明、暗、同、不同

2. 关键点分析

根据Halcon帮助文档种给出的 dyn_threshold的分割阈值的计算方式 
g_o=g_[OrigImage] 
g_t=g_[ThresholdImage] 
当选择light模式 
阈值计算公式: 
for LightDark = ‘light’ is: 
g_o >= g_t + Offset 
即为原图中大于等于g_t + Offset点的像素值被选中。

当选择dark 模式 
For LightDark = ‘dark’ the condition is: 
g_o <= g_t - Offset 
即为原图中小于等于g_t - Offset点的像素值被选中。

当选择equal 模式时: 
For LightDark = ‘equal’ it is: 
g_t - Offset <= g_o <= g_t + Offset 
原图中像素值在g_t - Offset 和 g_t + Offset之间的像素点被选中。

当选择not_equal模式时 
Finally, for LightDark = ‘not_equal’ it is: 
g_t - Offset > g_o or g_o > g_t + Offset 
原图中像素值在g_t - Offset 和 g_t + Offset之外的像素点被选中。 
以上是我对dyn_threshold算子的理解,欢迎大家批评指正

``` dev_update_off() * 参数定义 GaussSize := 5 ScratchLengthMin := 30 StainAreaMin := 50 * 获取图像文件列表 ImageFolder := 'E:/毕业论文/瓶盖图/' list_files(ImageFolder, ['files','follow_links'], ImageFiles) tuple_regexp_select(ImageFiles, ['\\.(tif|tiff|gif|bmp|jpg|jpeg|jp2|png|pcx|pgm|ppm|pbm|xwd|ima|hobj)$','ignore_case'], ImageFiles) * 创建日志文件 open_file('defect_log.txt', 'output', FileHandle) * 主循环 for ImgIndex := 0 to |ImageFiles| - 1 by 1 * 窗口管理 dev_close_window() dev_open_window(0, 0, 600, 400, 'black', WindowHandle) * 文件读取(带异常捕获) try read_image(Image, ImageFiles[ImgIndex]) catch (Exception) write_string(FileHandle, 'Failed:'+ImageFiles[ImgIndex]+'\n') continue endtry get_image_size (Image, Width, Height) dev_open_window (0, 0, 600, 400, 'black', WindowHandle) set_display_font (WindowHandle, 16, 'mono', 'true', 'false') dev_set_draw ('margin') dev_set_line_width (2) * 1. 图像预处理(保留V通道信息) try decompose3 (Image, R, G, B) trans_from_rgb (R, G, B, H, S, V, 'hsv') gauss_filter (V, VFiltered, GaussSize) catch (Exception) disp_message (WindowHandle, '预处理错误', 'window', 12, 12, 'red', 'true') stop() endtry * 2. 瓶盖区域定位 binary_threshold (VFiltered, Regions, 'max_separability', 'dark', UsedThreshold) closing_circle (Regions, RegionClosing, 5.5) connection (RegionClosing, ConnectedRegions) ImageArea := Width * Height MinArea := 0.3 * ImageArea MaxArea := 0.9 * ImageArea select_shape (ConnectedRegions, SelectedRegions, 'area', 'and', MinArea, MaxArea) select_shape (SelectedRegions, CapCandidates, ['circularity','compactness'], 'and', [0.6,0.5], [1,1]) fill_up (CapCandidates, RegionFillUp) shape_trans (RegionFillUp, CapRegion, 'convex') * 3. 缺陷检测流程 * 3.1 划痕检测 reduce_domain (VFiltered, CapRegion, VReduced) equ_histo_image (VReduced, VEquHist) edges_sub_pix (VEquHist, Edges, 'canny', 0.5, 15, 25) select_shape_xld (Edges, Scratches, ['contlength','width'], 'and', [ScratchLengthMin,1], [9999,3]) gen_region_contour_xld (Scratches, ScratchesRegion, 'filled') * 3.2 污渍检测 mean_image (VReduced, ImageMean, 31, 31) dyn_threshold (VReduced, ImageMean, Stains, 15, 'dark') connection (Stains, ConnectedStains) select_shape (ConnectedStains, SelectedStains, ['area','compactness'], 'and', [StainAreaMin,0.2], [9999,1.5]) * 4. 合并缺陷区域 gen_empty_obj (AllDefects) union2 (ScratchesRegion, SelectedStains, AllDefects) * 5. 后处理 connection (AllDefects, ConnectedDefects) closing_circle(ConnectedDefects, FinalDefects, 1.5) * 6. 结果显示 count_obj (FinalDefects, NumDefects) // 此时NumDefects会被正确初始化 dev_display (Image) dev_set_color ('green') dev_display (CapRegion) if (NumDefects > 0) * 划痕显示(蓝色) dev_set_color ('blue') dev_display (ScratchesRegion) * 污渍显示(紫色) dev_set_color ('magenta') dev_display (SelectedStains) * 添加图例说明 disp_message (WindowHandle, '检测结果:', 'window', 12, 12, 'white', 'true') disp_message (WindowHandle, '蓝色:划痕缺陷', 'window', 40, 12, 'blue', 'true') disp_message (WindowHandle, '紫色:污渍缺陷', 'window', 60, 12, 'magenta', 'true') else disp_message (WindowHandle, '检测通过:无缺陷', 'window', 12, 12, 'green', 'true') endif endfor *主循环结束```结果显示无缺陷,但过程识别到了划痕污渍
03-19
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值