POI 模板详解

概述

POI（Point of Interest，兴趣点）模板组是ColorVision中最复杂的模板系统之一，用于图像中关键点的检测、分析、校准和输出。POI系统广泛应用于：

显示设备像素级分析
光学标定
图像配准
质量检测
特征提取

模块组成

POI模板组包含6个子模块：

POI/
├── AlgorithmImp/      # POI算法实现
├── BuildPoi/          # POI构建
├── POIFilters/        # POI过滤器
├── POIGenCali/        # POI生成校准
├── POIOutput/         # POI输出
└── POIRevise/         # POI修正

技术特点

多阶段处理: 检测→过滤→校准→修正→输出
高度可配置: 每个阶段都有独立的参数模板
滤波器链: 支持多个过滤器串联
灵活输出: 支持多种数据格式

POI核心概念

POI数据结构

csharp

public class POIPoint
{
    public int Id { get; set; }              // POI唯一标识
    public double X { get; set; }            // X坐标
    public double Y { get; set; }            // Y坐标
    public PointType Type { get; set; }      // 点类型
    public double Value { get; set; }        // 关联数值
    public string Name { get; set; }         // 点名称
    public Dictionary\<string, object> Properties { get; set; }  // 扩展属性
}

POI类型

类型	说明	应用场景
Feature	特征点	角点、边缘点检测
Calibration	校准点	相机标定、畸变校正
Measurement	测量点	尺寸测量、定位
Defect	缺陷点	质量检测、异常识别
Reference	参考点	基准点、锚点

POI模板组织结构

1. AlgorithmImp - POI算法实现

职责: 实现各种POI检测算法

主要算法:

角点检测: Harris、Shi-Tomasi、FAST
特征点检测: SIFT、SURF、ORB
边缘检测: Canny、Sobel
圆形检测: Hough Circle
自定义检测: 基于模板匹配、颜色阈值等

参数示例:

csharp

public class POIDetectionParam : ParamModBase
{
    [Category("检测算法")]
    [Description("算法类型")]
    public POIAlgorithmType AlgorithmType { get; set; }
    
    [Category("检测参数")]
    [Description("质量阈值")]
    public double QualityThreshold { get; set; } = 0.01;
    
    [Category("检测参数")]
    [Description("最小距离")]
    public double MinDistance { get; set; } = 10.0;
    
    [Category("检测参数")]
    [Description("最大点数")]
    public int MaxPoints { get; set; } = 1000;
}

2. BuildPoi - POI构建

职责: 从检测结果构建POI数据结构

功能:

POI点聚类
POI层级组织
POI属性赋值
POI索引建立

参数示例:

csharp

public class POIBuildParam : ParamModBase
{
    [Category("构建参数")]
    [Description("聚类半径")]
    public double ClusterRadius { get; set; } = 5.0;
    
    [Category("构建参数")]
    [Description("最小聚类点数")]
    public int MinClusterSize { get; set; } = 3;
    
    [Category("输出设置")]
    [Description("POI命名模式")]
    public string NamingPattern { get; set; } = "POI_{0:D4}";
}

3. POIFilters - POI过滤器

职责: 过滤和筛选POI点

过滤器类型:

位置过滤器

csharp

public class POIPositionFilter : IPOIFilter
{
    public Rect ValidRegion { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        return points.Where(p => ValidRegion.Contains(new Point(p.X, p.Y))).ToList();
    }
}

值过滤器

csharp

public class POIValueFilter : IPOIFilter
{
    public double MinValue { get; set; }
    public double MaxValue { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        return points.Where(p => p.Value >= MinValue && p.Value <= MaxValue).ToList();
    }
}

距离过滤器

csharp

public class POIDistanceFilter : IPOIFilter
{
    public double MinDistance { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        // 移除距离过近的点
        var filtered = new List\<POIPoint>();
        foreach (var point in points)
        {
            if (!filtered.Any(p => Distance(p, point) < MinDistance))
            {
                filtered.Add(point);
            }
        }
        return filtered;
    }
}

参数示例:

csharp

public class POIFilterParam : ParamModBase
{
    [Category("位置过滤")]
    [Description("最小X坐标")]
    public double MinX { get; set; }
    
    [Category("位置过滤")]
    [Description("最大X坐标")]
    public double MaxX { get; set; }
    
    [Category("值过滤")]
    [Description("最小值")]
    public double MinValue { get; set; }
    
    [Category("值过滤")]
    [Description("最大值")]
    public double MaxValue { get; set; }
}

4. POIGenCali - POI生成校准

职责: 使用POI进行几何校准

功能:

相机内参标定
畸变系数计算
坐标系变换
刚体变换估计

参数示例:

csharp

public class POICalibrationParam : ParamModBase
{
    [Category("标定参数")]
    [Description("标定板类型")]
    public CalibrationBoardType BoardType { get; set; }
    
    [Category("标定参数")]
    [Description("标定板宽度(mm)")]
    public double BoardWidth { get; set; }
    
    [Category("标定参数")]
    [Description("标定板高度(mm)")]
    public double BoardHeight { get; set; }
    
    [Category("标定参数")]
    [Description("网格间距(mm)")]
    public double GridSpacing { get; set; }
}

标定流程:

5. POIOutput - POI输出

职责: 将POI数据输出为各种格式

支持格式:

JSON: 结构化数据
CSV: 表格数据
XML: 标记语言
Binary: 二进制格式（高效）
Image: 可视化图像

参数示例:

csharp

public class POIOutputParam : ParamModBase
{
    [Category("输出设置")]
    [Description("输出格式")]
    public OutputFormat Format { get; set; }
    
    [Category("输出设置")]
    [Description("输出路径")]
    public string OutputPath { get; set; }
    
    [Category("输出内容")]
    [Description("包含属性")]
    public bool IncludeProperties { get; set; } = true;
    
    [Category("输出内容")]
    [Description("包含图像")]
    public bool IncludeImage { get; set; } = false;
}

JSON输出示例:

json

{
  "version": "1.0",
  "timestamp": "2025-10-12T20:54:33Z",
  "points": [
    {
      "id": 1,
      "x": 123.45,
      "y": 678.90,
      "type": "Feature",
      "value": 0.95,
      "name": "POI_0001",
      "properties": {
        "quality": 0.98,
        "response": 125.3
      }
    }
  ]
}

6. POIRevise - POI修正

职责: 对检测到的POI进行后处理修正

修正类型:

亚像素精化: 提高定位精度
异常点剔除: 移除噪声点
坐标变换: 应用几何变换
值校正: 应用校准曲线

参数示例:

csharp

public class POIReviseParam : ParamModBase
{
    [Category("精化参数")]
    [Description("启用亚像素精化")]
    public bool EnableSubPixel { get; set; } = true;
    
    [Category("精化参数")]
    [Description("精化窗口大小")]
    public int RefineWindowSize { get; set; } = 5;
    
    [Category("异常检测")]
    [Description("异常值标准差倍数")]
    public double OutlierStdDev { get; set; } = 3.0;
}

POI处理流程

完整处理管道

典型工作流

csharp

public async Task<List\<POIPoint>> ProcessPOIPipeline(Mat image)
{
    // 1. POI检测
    var detector = new POIDetector();
    var rawPoints = detector.Detect(image, detectionParam);
    
    // 2. POI构建
    var builder = new POIBuilder();
    var structuredPoints = builder.Build(rawPoints, buildParam);
    
    // 3. POI过滤
    var filter = new POIFilterChain();
    filter.AddFilter(new POIPositionFilter());
    filter.AddFilter(new POIValueFilter());
    var filteredPoints = filter.Apply(structuredPoints, filterParam);
    
    // 4. POI校准（可选）
    if (needCalibration)
    {
        var calibrator = new POICalibrator();
        filteredPoints = calibrator.Calibrate(filteredPoints, calibrationParam);
    }
    
    // 5. POI修正
    var reviser = new POIReviser();
    var refinedPoints = reviser.Revise(filteredPoints, reviseParam);
    
    // 6. POI输出
    var outputer = new POIOutputer();
    await outputer.Save(refinedPoints, outputParam);
    
    return refinedPoints;
}

各子模块详解

AlgorithmImp - 算法实现

角点检测算法

Harris角点检测:

csharp

public class HarrisCornerDetector : IPOIAlgorithm
{
    public double Quality { get; set; } = 0.01;
    public double MinDistance { get; set; } = 10.0;
    public int BlockSize { get; set; } = 3;
    public double K { get; set; } = 0.04;
    
    public List\<POIPoint> Detect(Mat image)
    {
        Mat corners = new Mat();
        Cv2.CornerHarris(image, corners, BlockSize, 3, K);
        
        // 非极大值抑制
        var points = ExtractLocalMaxima(corners, Quality, MinDistance);
        return points;
    }
}

Shi-Tomasi角点检测:

csharp

public class ShiTomasiDetector : IPOIAlgorithm
{
    public int MaxCorners { get; set; } = 1000;
    public double QualityLevel { get; set; } = 0.01;
    public double MinDistance { get; set; } = 10.0;
    
    public List\<POIPoint> Detect(Mat image)
    {
        var corners = Cv2.GoodFeaturesToTrack(
            image, MaxCorners, QualityLevel, MinDistance);
        return ConvertToPOI(corners);
    }
}

BuildPoi - POI构建器

聚类构建

csharp

public class POIClusterBuilder : IPOIBuilder
{
    public double ClusterRadius { get; set; }
    public int MinClusterSize { get; set; }
    
    public List\<POIPoint> Build(List\<POIPoint> rawPoints)
    {
        var clusters = new List<List\<POIPoint>>();
        var visited = new HashSet\<int>();
        
        foreach (var point in rawPoints)
        {
            if (visited.Contains(point.Id))
                continue;
            
            var cluster = FindNeighbors(point, rawPoints, ClusterRadius);
            if (cluster.Count >= MinClusterSize)
            {
                clusters.Add(cluster);
                foreach (var p in cluster)
                    visited.Add(p.Id);
            }
        }
        
        // 计算聚类中心
        return clusters.Select(c => GetClusterCenter(c)).ToList();
    }
}

网格构建

csharp

public class POIGridBuilder : IPOIBuilder
{
    public int GridRows { get; set; }
    public int GridColumns { get; set; }
    public double GridSpacing { get; set; }
    
    public List\<POIPoint> Build(List\<POIPoint> rawPoints)
    {
        var grid = new POIPoint[GridRows, GridColumns];
        
        // 将检测点映射到网格
        foreach (var point in rawPoints)
        {
            var (row, col) = FindNearestGridCell(point);
            if (grid[row, col] == null || 
                point.Value > grid[row, col].Value)
            {
                grid[row, col] = point;
            }
        }
        
        return GridToList(grid);
    }
}

POIFilters - 过滤器系统

过滤器接口

csharp

public interface IPOIFilter
{
    string Name { get; }
    List\<POIPoint> Filter(List\<POIPoint> points);
}

链式过滤

csharp

public class POIFilterChain
{
    private List\<IPOIFilter> filters = new();
    
    public void AddFilter(IPOIFilter filter)
    {
        filters.Add(filter);
    }
    
    public List\<POIPoint> Apply(List\<POIPoint> points, POIFilterParam param)
    {
        var result = points;
        foreach (var filter in filters)
        {
            result = filter.Filter(result);
        }
        return result;
    }
}

内置过滤器

边界过滤器:

csharp

public class POIBoundaryFilter : IPOIFilter
{
    public string Name => "边界过滤";
    public Rect Boundary { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        return points.Where(p => 
            p.X >= Boundary.Left && p.X <= Boundary.Right &&
            p.Y >= Boundary.Top && p.Y <= Boundary.Bottom
        ).ToList();
    }
}

质量过滤器:

csharp

public class POIQualityFilter : IPOIFilter
{
    public string Name => "质量过滤";
    public double MinQuality { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        return points.Where(p => p.Value >= MinQuality).ToList();
    }
}

稀疏化过滤器:

csharp

public class POISparsificationFilter : IPOIFilter
{
    public string Name => "稀疏化";
    public double MinSpacing { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        var result = new List\<POIPoint>();
        var sortedPoints = points.OrderByDescending(p => p.Value).ToList();
        
        foreach (var point in sortedPoints)
        {
            if (!result.Any(p => Distance(p, point) < MinSpacing))
            {
                result.Add(point);
            }
        }
        return result;
    }
}

POIGenCali - 生成校准

相机标定

csharp

public class POICameraCalibration
{
    public CalibrationResult Calibrate(
        List<List\<POIPoint>> imagePoints,    // 图像点（多组）
        List<List\<Point3D>> objectPoints,    // 物理点（多组）
        Size imageSize)
    {
        // 标定相机内参
        var cameraMatrix = new Mat();
        var distCoeffs = new Mat();
        
        Cv2.CalibrateCamera(
            objectPoints.ToArray(),
            imagePoints.ToArray(),
            imageSize,
            cameraMatrix,
            distCoeffs,
            out var rvecs,
            out var tvecs
        );
        
        return new CalibrationResult
        {
            CameraMatrix = cameraMatrix,
            DistortionCoefficients = distCoeffs,
            RotationVectors = rvecs,
            TranslationVectors = tvecs
        };
    }
}

平面变换

csharp

public class POIPerspectiveTransform
{
    public Mat GetTransformMatrix(
        List\<POIPoint> sourcePoints,
        List\<POIPoint> targetPoints)
    {
        var src = sourcePoints.Select(p => new Point2f((float)p.X, (float)p.Y)).ToArray();
        var dst = targetPoints.Select(p => new Point2f((float)p.X, (float)p.Y)).ToArray();
        
        return Cv2.GetPerspectiveTransform(src, dst);
    }
}

POIRevise - POI修正

亚像素精化

csharp

public class POISubPixelRefine
{
    public int WindowSize { get; set; } = 5;
    public int MaxIterations { get; set; } = 30;
    public double Epsilon { get; set; } = 0.001;
    
    public List\<POIPoint> Refine(Mat image, List\<POIPoint> points)
    {
        var corners = points.Select(p => new Point2f((float)p.X, (float)p.Y)).ToArray();
        
        var criteria = new TermCriteria(
            CriteriaTypes.Eps | CriteriaTypes.MaxIter,
            MaxIterations,
            Epsilon
        );
        
        Cv2.CornerSubPix(
            image,
            corners,
            new Size(WindowSize, WindowSize),
            new Size(-1, -1),
            criteria
        );
        
        // 更新点坐标
        for (int i = 0; i < points.Count; i++)
        {
            points[i].X = corners[i].X;
            points[i].Y = corners[i].Y;
        }
        
        return points;
    }
}

异常点剔除

csharp

public class POIOutlierRemoval
{
    public double StdDevMultiplier { get; set; } = 3.0;
    
    public List\<POIPoint> RemoveOutliers(List\<POIPoint> points)
    {
        // 基于统计方法移除异常点
        var values = points.Select(p => p.Value).ToArray();
        var mean = values.Average();
        var stdDev = CalculateStdDev(values, mean);
        
        return points.Where(p => 
            Math.Abs(p.Value - mean) <= StdDevMultiplier * stdDev
        ).ToList();
    }
}

POIOutput - 输出实现

JSON输出

csharp

public class POIJsonOutput : IPOIOutput
{
    public void Save(List\<POIPoint> points, string filePath)
    {
        var output = new
        {
            Version = "1.0",
            Timestamp = DateTime.Now,
            Count = points.Count,
            Points = points.Select(p => new
            {
                p.Id,
                p.X,
                p.Y,
                Type = p.Type.ToString(),
                p.Value,
                p.Name,
                p.Properties
            })
        };
        
        var json = JsonConvert.SerializeObject(output, Formatting.Indented);
        File.WriteAllText(filePath, json, Encoding.UTF8);
    }
}

CSV输出

csharp

public class POICsvOutput : IPOIOutput
{
    public void Save(List\<POIPoint> points, string filePath)
    {
        var csv = new StringBuilder();
        csv.AppendLine("ID,X,Y,Type,Value,Name");
        
        foreach (var point in points)
        {
            csv.AppendLine($"{point.Id},{point.X},{point.Y},{point.Type},{point.Value},{point.Name}");
        }
        
        File.WriteAllText(filePath, csv.ToString(), Encoding.UTF8);
    }
}

使用指南

快速开始

csharp

// 1. 创建POI模板
var templatePOI = new TemplatePoi();
templatePOI.Load();

// 2. 配置检测参数
var detectionParam = new POIDetectionParam
{
    AlgorithmType = POIAlgorithmType.ShiTomasi,
    QualityThreshold = 0.01,
    MinDistance = 10.0,
    MaxPoints = 500
};

// 3. 配置过滤参数
var filterParam = new POIFilterParam
{
    MinX = 0,
    MaxX = 1920,
    MinY = 0,
    MaxY = 1080,
    MinValue = 0.1
};

// 4. 执行POI检测
var algorithm = new AlgorithmPOI();
var result = await algorithm.ExecuteAsync(image, detectionParam);

// 5. 应用过滤
var filtered = ApplyFilters(result.Points, filterParam);

// 6. 输出结果
var output = new POIJsonOutput();
output.Save(filtered, "poi_result.json");

高级用法

自定义POI算法

csharp

public class CustomPOIDetector : IPOIAlgorithm
{
    public List\<POIPoint> Detect(Mat image)
    {
        var points = new List\<POIPoint>();
        
        // 自定义检测逻辑
        // 例如：基于颜色的检测
        var hsv = new Mat();
        Cv2.CvtColor(image, hsv, ColorConversionCodes.BGR2HSV);
        
        var mask = new Mat();
        Cv2.InRange(hsv, new Scalar(100, 50, 50), new Scalar(130, 255, 255), mask);
        
        var contours = Cv2.FindContoursAsArray(mask, RetrievalModes.External, ContourApproximationModes.ApproxSimple);
        
        foreach (var contour in contours)
        {
            var moments = Cv2.Moments(contour);
            var center = new Point2d(
                moments.M10 / moments.M00,
                moments.M01 / moments.M00
            );
            
            points.Add(new POIPoint
            {
                X = center.X,
                Y = center.Y,
                Type = PointType.Feature,
                Value = Cv2.ContourArea(contour)
            });
        }
        
        return points;
    }
}

自定义过滤器

csharp

public class CustomPOIFilter : IPOIFilter
{
    public string Name => "自定义过滤";
    
    // 自定义条件
    public Func\<POIPoint, bool> Condition { get; set; }
    
    public List\<POIPoint> Filter(List\<POIPoint> points)
    {
        return points.Where(p => Condition(p)).ToList();
    }
}

// 使用
var filter = new CustomPOIFilter
{
    Condition = p => p.X > 100 && p.Y > 100 && p.Value > 0.5
};

最佳实践

1. 检测参数选择

csharp

// 高质量检测（精确，较慢）
var highQualityParam = new POIDetectionParam
{
    AlgorithmType = POIAlgorithmType.ShiTomasi,
    QualityThreshold = 0.001,  // 低阈值，检测更多点
    MinDistance = 5.0,          // 允许密集点
    MaxPoints = 2000
};

// 快速检测（快速，一般精度）
var fastParam = new POIDetectionParam
{
    AlgorithmType = POIAlgorithmType.FAST,
    QualityThreshold = 0.05,   // 高阈值，检测少量点
    MinDistance = 20.0,         // 较大间距
    MaxPoints = 200
};

2. 过滤器配置

csharp

// 建议的过滤器顺序
var filterChain = new POIFilterChain();

// 1. 首先过滤边界外的点
filterChain.AddFilter(new POIBoundaryFilter { Boundary = validRegion });

// 2. 然后过滤低质量点
filterChain.AddFilter(new POIQualityFilter { MinQuality = 0.1 });

// 3. 最后稀疏化
filterChain.AddFilter(new POISparsificationFilter { MinSpacing = 15.0 });

3. 性能优化

csharp

// 使用图像金字塔加速检测
public List\<POIPoint> DetectMultiScale(Mat image)
{
    var pyramid = BuildPyramid(image, 3);  // 3层金字塔
    var points = new List\<POIPoint>();
    
    for (int level = 0; level < pyramid.Count; level++)
    {
        var scale = Math.Pow(2, level);
        var levelPoints = detector.Detect(pyramid[level]);
        
        // 转换回原始尺度
        foreach (var p in levelPoints)
        {
            p.X *= scale;
            p.Y *= scale;
            points.Add(p);
        }
    }
    
    return points;
}

4. 结果验证

csharp

public bool ValidatePOIResult(List\<POIPoint> points)
{
    // 检查点数
    if (points.Count == 0)
    {
        Console.WriteLine("警告: 未检测到POI点");
        return false;
    }
    
    // 检查分布
    var bounds = GetBoundingBox(points);
    var coverage = (bounds.Width * bounds.Height) / (imageWidth * imageHeight);
    if (coverage < 0.1)
    {
        Console.WriteLine("警告: POI分布过于集中");
        return false;
    }
    
    // 检查密度
    var density = points.Count / (bounds.Width * bounds.Height);
    if (density > 0.01)
    {
        Console.WriteLine("警告: POI过于密集，可能有噪声");
        return false;
    }
    
    return true;
}

5. 错误处理

csharp

public async Task\<POIResult> SafeDetectPOI(Mat image, POIDetectionParam param)
{
    try
    {
        // 验证输入
        if (image == null || image.Empty())
            throw new ArgumentException("图像为空");
        
        if (param == null)
            throw new ArgumentNullException(nameof(param));
        
        // 执行检测
        var points = detector.Detect(image, param);
        
        // 验证结果
        if (!ValidatePOIResult(points))
        {
            return new POIResult 
            { 
                Success = false, 
                Message = "检测结果不符合预期" 
            };
        }
        
        return new POIResult 
        { 
            Success = true, 
            Points = points 
        };
    }
    catch (Exception ex)
    {
        log.Error($"POI检测失败: {ex.Message}", ex);
        return new POIResult 
        { 
            Success = false, 
            Message = ex.Message 
        };
    }
}

POI 模板详解 ​

目录 ​

概述 ​

模块组成 ​

技术特点 ​

POI核心概念 ​

POI数据结构 ​

POI类型 ​

POI模板组织结构 ​

1. AlgorithmImp - POI算法实现 ​

2. BuildPoi - POI构建 ​

3. POIFilters - POI过滤器 ​

位置过滤器 ​

值过滤器 ​

距离过滤器 ​

4. POIGenCali - POI生成校准 ​

5. POIOutput - POI输出 ​

6. POIRevise - POI修正 ​

POI处理流程 ​

完整处理管道 ​

典型工作流 ​

各子模块详解 ​

AlgorithmImp - 算法实现 ​

角点检测算法 ​

BuildPoi - POI构建器 ​

聚类构建 ​

网格构建 ​

POIFilters - 过滤器系统 ​

过滤器接口 ​

链式过滤 ​

内置过滤器 ​

POIGenCali - 生成校准 ​

相机标定 ​

平面变换 ​

POIRevise - POI修正 ​

亚像素精化 ​

异常点剔除 ​

POIOutput - 输出实现 ​

JSON输出 ​

CSV输出 ​

使用指南 ​

快速开始 ​

高级用法 ​

自定义POI算法 ​

自定义过滤器 ​

最佳实践 ​

1. 检测参数选择 ​

2. 过滤器配置 ​

3. 性能优化 ​

4. 结果验证 ​

5. 错误处理 ​

相关资源 ​

POI 模板详解

目录

概述

模块组成

技术特点

POI核心概念

POI数据结构

POI类型

POI模板组织结构

1. AlgorithmImp - POI算法实现

2. BuildPoi - POI构建

3. POIFilters - POI过滤器

位置过滤器

值过滤器

距离过滤器

4. POIGenCali - POI生成校准

5. POIOutput - POI输出

6. POIRevise - POI修正

POI处理流程

完整处理管道

典型工作流

各子模块详解

AlgorithmImp - 算法实现

角点检测算法

BuildPoi - POI构建器

聚类构建

网格构建

POIFilters - 过滤器系统

过滤器接口

链式过滤

内置过滤器

POIGenCali - 生成校准

相机标定

平面变换

POIRevise - POI修正

亚像素精化

异常点剔除

POIOutput - 输出实现

JSON输出

CSV输出

使用指南

快速开始

高级用法

自定义POI算法

自定义过滤器

最佳实践

1. 检测参数选择

2. 过滤器配置

3. 性能优化

4. 结果验证

5. 错误处理

相关资源