应用阈值后无法显示黑白图像 求HSL车道检测代码Bug修复

我正在开展基于HSL图像格式阈值的车道线检测项目，编写了如下代码。代码无报错，但输出仅为全黑图像。我已尝试调整阈值参数、更换多张图像，问题仍未解决，恳请帮忙修复Bug。

原始图像：包含道路、车道线及桥洞阴影的道路场景图
当前输出：全黑图像

代码 #

import cv2
import numpy as np
import logging
import matplotlib.pyplot as plt

logging.basicConfig(level=logging.INFO, format='%(funcName)s:%(message)s')


class LaneFinding:
    def __init__(self, img):
        self.img = img

    def get_hsl_transform(self, thresh: tuple):
        """
        This method will return a HSL image for the given image for which the pixel values are adjusted between the 
        lower and higher threshold.

        The values between lower and upper threshold will be 1 and all other values will be zero.
        
        :param tuple thresh: Tuple of two values, the lower value threshold and upper threshold for the image.
        
        """
        # Convert the given RGB image to HSL image
        img_hsl = cv2.cvtColor(self.img, cv2.COLOR_RGB2HLS)

        # Separate individula channels in the imge and create zero masks for each channel.
        h = img_hsl[:,:,0]
        l = img_hsl[:,:,1]
        s = img_hsl[:,:,2]

        binary_h = np.zeros_like(img_hsl)
        binary_s = np.zeros_like(img_hsl)
        binary_l = np.zeros_like(img_hsl)

        # Apply threshold to individual channels and make a binary image.
        binary_h[(h>thresh[0]) & (h<=thresh[1])] = 1
        binary_s[(s>thresh[0]) & (s<=thresh[1])] = 1
        binary_l[(l>thresh[0]) & (l<=thresh[1])] = 1

        return binary_h, binary_s, binary_l
    



if __name__ == '__main__':

    # This will only check if the new methods added in the class above work or not. 

    img = cv2.imread(r"Gradient_and_color_space\\bridge_shadow.jpg")
    thresh = (10, 100)

    lane_finding = LaneFinding(img)

    _, binary_s, _ = lane_finding.get_hsl_transform(thresh=thresh)
    plt.imshow(binary_s, cmap='gray')
    plt.show()

代码存在三个核心问题：

1. 颜色空间转换错误 #

cv2.imread读取图像默认返回BGR格式，但代码中使用cv2.COLOR_RGB2HLS进行转换，这会导致颜色通道映射错误，HSL转换结果完全偏离预期。

2. 二进制掩码维度不匹配 #

分离出的h/s/l是单通道（2D数组），但创建的binary_h/binary_s/binary_l用np.zeros_like(img_hsl)生成了三通道（3D数组）。赋值1时仅会给单通道赋值，而plt.imshow显示三通道数组时，单通道的1会被视为极低亮度，最终呈现全黑。

3. 阈值参数不符合S通道特征 #

原始图像中车道线的饱和度（S通道）值较高，原阈值(10,100)过低，无法筛选出车道线像素。

修复后的代码 #

import cv2
import numpy as np
import logging
import matplotlib.pyplot as plt

logging.basicConfig(level=logging.INFO, format='%(funcName)s:%(message)s')


class LaneFinding:
    def __init__(self, img):
        self.img = img

    def get_hsl_transform(self, thresh: tuple):
        # 将BGR图像转换为HSL格式
        img_hsl = cv2.cvtColor(self.img, cv2.COLOR_BGR2HLS)

        h = img_hsl[:,:,0]
        l = img_hsl[:,:,1]
        s = img_hsl[:,:,2]

        # 创建单通道的二进制掩码，与分离出的通道维度一致
        binary_h = np.zeros_like(h)
        binary_s = np.zeros_like(s)
        binary_l = np.zeros_like(l)

        # 对各通道应用阈值，用255作为高亮值更符合图像显示习惯
        binary_h[(h > thresh[0]) & (h <= thresh[1])] = 255
        binary_s[(s > thresh[0]) & (s <= thresh[1])] = 255
        binary_l[(l > thresh[0]) & (l <= thresh[1])] = 255

        return binary_h, binary_s, binary_l
    



if __name__ == '__main__':
    img = cv2.imread(r"Gradient_and_color_space\\bridge_shadow.jpg")
    # 针对S通道调整阈值，适配车道线的饱和度特征
    thresh = (90, 255)

    lane_finding = LaneFinding(img)

    _, binary_s, _ = lane_finding.get_hsl_transform(thresh=thresh)
    plt.imshow(binary_s, cmap='gray')
    plt.show()

修复说明 #

• 把cv2.COLOR_RGB2HLS改为cv2.COLOR_BGR2HLS，匹配cv2.imread的BGR输入格式；
• 用np.zeros_like(h)创建单通道掩码，确保维度匹配；
• 将赋值从1改为255，因为plt.imshow的灰度图中255对应白色，更直观；
• 调整S通道阈值为(90,255)，可以有效筛选出高饱和度的车道线像素。

内容的提问来源于stack exchange，提问作者programmer_04_03