如何用openCV在这张图片上准确检测brown/black/grey/white

How to accurately detect brown/black/grey/white on this picture with openCV

我首先使用 HoughCircles 找到每个圆的中心,但后来我意识到我还需要知道找到的相应圆的颜色,所以我尝试了另一种方法(见下文)。

注意:圈子会随机放置,没有硬编码。

这是图片的样子:

[![在此处输入图片描述][1]][1]

问题:

很难获得准确的 HSV 值来正确检测标题中的颜色,而且图像质量也不是最好的。我认为那些冰球中间的圆圈是为了帮助我们区分它们,但由于大多数圆盘中都有一个浅蓝色圆圈,我不确定它有什么帮助哈哈。

我尝试了什么:

1.

我使用 openCV trackbar 来获得每种颜色(除了提到的那些)的近似下限和上限,这真的很难获得。

2.

我把蒙版贴在图片上,然后用力矩找到圆心

    import cv2
    import numpy as np
    
    img = cv2.imread('Photos/lastBoard.png')
    frame_hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    #red color
    lower_values = np.array([0,123,40])
    upper_values = np.array([5,255,114])
    
    mask = cv2.inRange(frame_hsv, lower_values, upper_values)
    
    contours, hierarchy = cv2.findContours(mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    for c in contours:
        M = cv2.moments(c)
        if M["m00"] != 0:
            #####
            (x,y),radius = cv2.minEnclosingCircle(c)
            center = (int(x),int(y))
            
            radius = int(radius)
            cX = int(M["m10"] / M["m00"])
    
            cY = int(M["m01"] / M["m00"])
            #this condition is just to tell to detect in that area of the image only
            if cX > 500 and radius >8:
                cv2.circle(img, (cX, cY), 2, (0,255,0), -1)
cv2.imshow("Image", img)

cv2.waitKey(0)

我需要什么帮助

找到棕色、黑色、白色和灰色真的很难,我的方法似乎不太准确。我有更好的方法吗?非常感谢

这是一种可行的方法,它仍然使用 HSV 颜色 space,您必须获得正确的 HSV 范围值。查找目标颜色的 RGB -> HSV 等效值。您肯定可以从一些 预处理 中获益,以更好地清理您的面具。您还可以实施 轮廓过滤器 ,因为您正在寻找的感兴趣的斑点(冰球)具有非常 不同的属性 ,例如,纵横比比率,面积,当然还有圆度。我建议采取以下步骤:

  1. 获取您正在寻找的每个目标冰球的 HSV 个值
  2. 定义 upperlower 范围值
  3. 阈值 HSV 图像以获得二进制掩码
  4. 应用区域过滤器去除小噪声
  5. 应用一些形态学 (Dilate+Erode) 来改进目标斑点
  6. 获取外轮廓(忽略内轮廓)
  7. 将这些轮廓转换为bounding rectangles
  8. 获取两个 bounding rectangles 属性:aspect ratioarea
  9. 根据阈值属性值过滤边界矩形

我们来看代码:

# importing cv2 and numpy:
import numpy as np
import cv2

# image path
path = "C://opencvImages//"
fileName = "board.png"

# Reading an image in default mode:
inputImage = cv2.imread(path + fileName)

# Convert the image to HSV:
frame_hsv = cv2.cvtColor(inputImage, cv2.COLOR_BGR2HSV)

# Prepare a dictionary to store the lower and upper
# HSV thresholds:
rangeDictionary = {}

# brown color
lower_values = np.array([6, 63, 0])
upper_values = np.array([23, 255, 81])

# push it into the dictionary:
rangeDictionary[0] = (lower_values, upper_values, "brown")

# gray color
lower_values = np.array([23, 0, 0])
upper_values = np.array([80, 105, 107])

# push it into the dictionary:
rangeDictionary[1] = (lower_values, upper_values, "gray")

# white color
lower_values = np.array([37, 0, 131])
upper_values = np.array([170, 25, 152])

# push it into the dictionary:
rangeDictionary[2] = (lower_values, upper_values, "white")

# Store results here:
targetRectangles = []

到目前为止,我有 个目标颜色的 HSV 等价物。我已经为这些颜色定义了 upperlower 阈值,并将它们存储在 dictionary 中。这个想法是遍历这个字典并相应地提取每个颜色范围:

# Loop through the dictionary and locate each circle:
for i in rangeDictionary:

    # Get current lower and upper range values:
    current_LowRange = rangeDictionary[i][0]
    current_UppRange = rangeDictionary[i][1]

    # Create the HSV mask
    mask = cv2.inRange(frame_hsv, current_LowRange, current_UppRange)

    # Run a minimum area filter:
    minArea = 800
    mask = areaFilter(minArea, mask)

对于第一种颜色,这是未过滤的二进制掩码:

你在这里看到我已经实现了 areaFilter。这将去除小于 800 的斑点,让我们开始以正确的方式清洁您的面具。这个函数定义在post的末尾。接下来是一些 morphology 来进一步定义目标 blob:

    # Pre-process mask:
    kernelSize = 3

    structuringElement = cv2.getStructuringElement(cv2.MORPH_RECT, (kernelSize, kernelSize))
    iterations = 10

    mask = cv2.morphologyEx(mask, cv2.MORPH_DILATE, structuringElement, None, None, iterations, cv2.BORDER_REFLECT101)
    mask = cv2.morphologyEx(mask, cv2.MORPH_ERODE, structuringElement, None, None, iterations, cv2.BORDER_REFLECT101)

这是过滤后的掩码:

很好,嗯?没什么特别的,只是一个非常激进的 dilation + erosion 链。我想定义冰球漂亮干净。根据输入图像的大小,您可能需要调整 iterations 值。让我们继续。接下来的步骤(仍在循环内)是计算 contours(仅外部的)并将每个 contour 近似为 polygon,然后近似为 rectangle

    # Find the big contours/blobs on the filtered image:
    contours, hierarchy = cv2.findContours(mask, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE)

    # List to store all the bounding rectangles:
    contours_poly = [None] * len(contours)
    boundRect = []

    # Alright, just look for the outer bounding boxes:
    for b, c in enumerate(contours):

        if hierarchy[0][b][3] == -1:

            # Approximate the contour to a polygon:
            contours_poly = cv2.approxPolyDP(c, 3, True)
            # Convert the polygon to a bounding rectangle:
            boundRect = cv2.boundingRect(contours_poly)

现在我们正在处理 bounding rectangles 并且操作变得更加简单。让我们获取矩形的尺寸并计算几个参数:aspectRatioarea。使用一些启发式方法,我已经设置了将用于过滤矩形的最小阈值:

            # Get the dimensions of the bounding rect:
            rectX = boundRect[0]
            rectY = boundRect[1]
            rectWidth = boundRect[2]
            rectHeight = boundRect[3]

            rectArea = rectWidth * rectHeight

            # Calculate the aspect ratio:
            aspectRatio = rectWidth / rectHeight
            delta = abs(1.0 - aspectRatio)

            # Set the min threshold values to identify the
            # blob of interest:
            minArea = 1000
            epsilon = 0.2

            # Is this bounding rectangle one the one we
            # are looking for?
            if rectArea > minArea and delta < epsilon:

                # Set a color:
                color = (0, 255, 0)
                inputCopy = inputImage.copy()

                # Draw the current rectangle on a copy of the BGR input:
                cv2.rectangle(inputCopy, (int(rectX), int(rectY)),
                              (int(rectX + rectWidth), int(rectY + rectHeight)), color, 2)
                # Store this bounding rectangle:
                targetRectangles.append(boundRect)


                # Label the current mask:
                currentColor = rangeDictionary[i][2]

                org = (rectX, rectY -10)
                font = cv2.FONT_HERSHEY_SIMPLEX
                color = (255, 0, 0)
                cv2.putText(inputCopy, currentColor, org, font,
                            0.5, color, 1, cv2.LINE_AA)

                cv2.imwrite(path + "colorMask_"+currentColor+".png", inputCopy)

我在输入的深层副本上另外绘制了目标矩形,并绘制了漂亮的文本来识别颜色,查看结果:

“B-b-但是伙计,黑色冰球怎么样?!” 好吧,我得给你留点事做。如果您一直遵循到现在,应该很容易获得额外的面具。这是areaFilter函数的定义和实现:

def areaFilter(minArea, inputImage):

    # Perform an area filter on the binary blobs:
    componentsNumber, labeledImage, componentStats, componentCentroids = \
    cv2.connectedComponentsWithStats(inputImage, connectivity=4)

    # Get the indices/labels of the remaining components based on the area stat
    # (skip the background component at index 0)
    remainingComponentLabels = [i for i in range(1, componentsNumber) if componentStats[i][4] >= minArea]

    # Filter the labeled pixels based on the remaining labels,
    # assign pixel intensity to 255 (uint8) for the remaining pixels
    filteredImage = np.where(np.isin(labeledImage, remainingComponentLabels) == True, 255, 0).astype('uint8')

    return filteredImage

该死,看看所有这些,我可能应该在您的项目报告中的某个地方得到承认。希望您发现此信息有用。