[已解决]多通道图片阈值分割

125404629

高光谱多通道图片阈值分割。

发生异常: error
OpenCV(4.7.0) D:\a\opencv-python\opencv-python\opencv\modules\core\src\arithm.cpp:230: error: (-215:Assertion failed) (mtype == CV_8U || mtype == CV_8S) && _mask.sameSize(*psrc1) in function 'cv::binary_op'
  File "D:\0000可见光2\程序\光谱提取\提取光谱.py", line 38, in <module>
    background_removed_image = cv2.bitwise_and(rawimg_RT, rawimg_RT, mask=binary_mask)
cv2.error: OpenCV(4.7.0) D:\a\opencv-python\opencv-python\opencv\modules\core\src\arithm.cpp:230: error: (-215:Assertion failed) (mtype == CV_8U || mtype == CV_8S) && _mask.sameSize(*psrc1) in function 'cv::binary_op'


输出结果：
(950, 384) (950, 384, 288)
uint8 float32
(950, 384, 288) (950, 384, 288)
uint8 uint8

代码：

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

# Load the hyperspectral image
rawimg_RT = np.fromfile('1-1_RT.raw', dtype=np.float32)
rawimg_RT = rawimg_RT.reshape(950, 384, 288)

# Select a suitable band with maximum contrast between peanuts and background
band_for_masking = 42  # You may need to adjust this based on your data

# Calculate the difference image between peanuts and background
difference_image = rawimg_RT[:,: , band_for_masking]

# Convert the difference image to an 8-bit image
normalized_difference = cv2.normalize(difference_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)

# Apply Otsu's thresholding to create a binary mask
#_, binary_mask = cv2.threshold(normalized_difference, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
_,binary_mask=cv2.threshold(normalized_difference,0.6,1,cv2.THRESH_BINARY)#阈值
print(binary_mask.shape,rawimg_RT.shape)
print(binary_mask.dtype,rawimg_RT.dtype)


cv2.imshow('Background Removed Image', binary_mask)
cv2.waitKey(0)
cv2.destroyAllWindows()


num_channels = 288  # 例如，创建一个3通道的掩膜
binary_mask = np.stack((binary_mask,) * num_channels, axis=2)
rawimg_RT = rawimg_RT.astype(np.uint8)
print(binary_mask.shape,rawimg_RT.shape)
print(binary_mask.dtype,rawimg_RT.dtype)
#binary_mask = np.array(binary_mask, dtype=np.uint8)

# 对高光谱图像应用二值化掩膜
background_removed_image = cv2.bitwise_and(rawimg_RT, rawimg_RT, mask=binary_mask)

# 显示去除背景后的图像
cv2.imshow('Background Removed Image', background_removed_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

最佳答案

月排行榜 / 总排行榜

Mike_python小

2023-11-3 18:25:13

根据您提供的代码和错误信息，我注意到问题可能出现在以下几个地方：

1. 数据类型不匹配：根据错误信息，_mask.sameSize(*psrc1)的断言失败，这可能是由于二值化掩膜（binary_mask）的数据类型与原始图像（rawimg_RT）的数据类型不匹配导致的。请确保它们具有相同的数据类型。

   在代码中，您可以尝试将以下行：

   rawimg_RT = rawimg_RT.astype(np.uint8)   

2. 通道数不匹配：在创建二值化掩膜后，您将其复制为与原始图像具有相同通道数的形状。但是，根据输出结果，binary_mask的形状为 (950, 384, 1)，而 rawimg_RT 的形状为 (950, 384, 288)。由于通道数不匹配，无法使用 cv2.bitwise_and 函数将它们应用于原始图像上。

   您可以尝试重新构建具有相同通道数的二值化掩膜，例如：

   num_channels = rawimg_RT.shape[2]   binary_mask = np.repeat(binary_mask, num_channels, axis=2)   

请参考修改后的代码如下所示：

import numpy as npimport matplotlib.pyplot as pltimport cv2# Load the hyperspectral imagerawimg_RT = np.fromfile('1-1_RT.raw', dtype=np.float32)rawimg_RT = rawimg_RT.reshape(950, 384, 288)# Select a suitable band with maximum contrast between peanuts and backgroundband_for_masking = 42  # You may need to adjust this based on your data# Calculate the difference image between peanuts and backgrounddifference_image = rawimg_RT[:,:,band_for_masking]# Convert the difference image to an 8-bit imagenormalized_difference = cv2.normalize(difference_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)# Apply Otsu's thresholding to create a binary mask_, binary_mask = cv2.threshold(normalized_difference, 0.6, 1, cv2.THRESH_BINARY)cv2.imshow('Binary Mask', binary_mask)cv2.waitKey(0)cv2.destroyAllWindows() # Match the number of channels in the binary mask to the original imagenum_channels = rawimg_RT.shape[2]binary_mask = np.repeat(binary_mask, num_channels, axis=2)# Convert data types to matchrawimg_RT = rawimg_RT.astype(np.uint8)binary_mask = binary_mask.astype(np.uint8)# Apply binary mask to the hyperspectral imagebackground_removed_image = cv2.bitwise_and(rawimg_RT, rawimg_RT, mask=binary_mask)cv2.imshow('Background Removed Image', background_removed_image)cv2.waitKey(0)cv2.destroyAllWindows()

请尝试使用修改后的代码，并确保 1-1_RT.raw 文件存在于脚本所在的目录中。如果问题仍然存在，请提供更多关于错误的详细信息，以便我们进一步排查。

球一个最佳答案谢谢啦！这对我非常重要！

跳转到最佳答案楼层

Mike_python小

根据您提供的代码和错误信息，我注意到问题可能出现在以下几个地方：

1. 数据类型不匹配：根据错误信息，_mask.sameSize(*psrc1)的断言失败，这可能是由于二值化掩膜（binary_mask）的数据类型与原始图像（rawimg_RT）的数据类型不匹配导致的。请确保它们具有相同的数据类型。

   在代码中，您可以尝试将以下行：

   rawimg_RT = rawimg_RT.astype(np.uint8)   

2. 通道数不匹配：在创建二值化掩膜后，您将其复制为与原始图像具有相同通道数的形状。但是，根据输出结果，binary_mask的形状为 (950, 384, 1)，而 rawimg_RT 的形状为 (950, 384, 288)。由于通道数不匹配，无法使用 cv2.bitwise_and 函数将它们应用于原始图像上。

   您可以尝试重新构建具有相同通道数的二值化掩膜，例如：

   num_channels = rawimg_RT.shape[2]   binary_mask = np.repeat(binary_mask, num_channels, axis=2)   

请参考修改后的代码如下所示：

import numpy as npimport matplotlib.pyplot as pltimport cv2# Load the hyperspectral imagerawimg_RT = np.fromfile('1-1_RT.raw', dtype=np.float32)rawimg_RT = rawimg_RT.reshape(950, 384, 288)# Select a suitable band with maximum contrast between peanuts and backgroundband_for_masking = 42  # You may need to adjust this based on your data# Calculate the difference image between peanuts and backgrounddifference_image = rawimg_RT[:,:,band_for_masking]# Convert the difference image to an 8-bit imagenormalized_difference = cv2.normalize(difference_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)# Apply Otsu's thresholding to create a binary mask_, binary_mask = cv2.threshold(normalized_difference, 0.6, 1, cv2.THRESH_BINARY)cv2.imshow('Binary Mask', binary_mask)cv2.waitKey(0)cv2.destroyAllWindows() # Match the number of channels in the binary mask to the original imagenum_channels = rawimg_RT.shape[2]binary_mask = np.repeat(binary_mask, num_channels, axis=2)# Convert data types to matchrawimg_RT = rawimg_RT.astype(np.uint8)binary_mask = binary_mask.astype(np.uint8)# Apply binary mask to the hyperspectral imagebackground_removed_image = cv2.bitwise_and(rawimg_RT, rawimg_RT, mask=binary_mask)cv2.imshow('Background Removed Image', background_removed_image)cv2.waitKey(0)cv2.destroyAllWindows()

请尝试使用修改后的代码，并确保 1-1_RT.raw 文件存在于脚本所在的目录中。如果问题仍然存在，请提供更多关于错误的详细信息，以便我们进一步排查。

球一个最佳答案谢谢啦！这对我非常重要！

isdkz

在代码中，出现了OpenCV的错误信息：Assertion failed，这是由于在进行二值化掩膜操作时出现了类型不匹配的错误。

错误的原因是在进行二值化操作时， normalized_difference  是一个  uint8  类型的图像，而  binary_mask  是一个  float32  类型的图像，这导致了类型不匹配的错误。

要解决这个问题，我们需要将  binary_mask  的数据类型转换为  uint8 ，以与  normalized_difference  的数据类型匹配。

在代码中，已经使用了  astype  函数将  rawimg_RT  的数据类型转换为  uint8 ，我们可以使用同样的方法将  binary_mask  的数据类型转换为  uint8 。

将  binary_mask = np.array(binary_mask, dtype=np.uint8)  这行代码取消注释，并将其修改为  binary_mask = binary_mask.astype(np.uint8) 。

修改后的代码如下：

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

# Load the hyperspectral image
rawimg_RT = np.fromfile('1-1_RT.raw', dtype=np.float32)
rawimg_RT = rawimg_RT.reshape(950, 384, 288)

# Select a suitable band with maximum contrast between peanuts and background
band_for_masking = 42  # You may need to adjust this based on your data

# Calculate the difference image between peanuts and background
difference_image = rawimg_RT[:,:, band_for_masking]

# Convert the difference image to an 8-bit image
normalized_difference = cv2.normalize(difference_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)

# Apply Otsu's thresholding to create a binary mask
#_, binary_mask = cv2.threshold(normalized_difference, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
_,binary_mask=cv2.threshold(normalized_difference,0.6,1,cv2.THRESH_BINARY)#阈值
print(binary_mask.shape,rawimg_RT.shape)
print(binary_mask.dtype,rawimg_RT.dtype)

cv2.imshow('Background Removed Image', binary_mask)
cv2.waitKey(0)
cv2.destroyAllWindows()

num_channels = 288  # 例如，创建一个3通道的掩膜
binary_mask = np.stack((binary_mask,) * num_channels, axis=2)
rawimg_RT = rawimg_RT.astype(np.uint8)
binary_mask = binary_mask.astype(np.uint8)

# 对高光谱图像应用二值化掩膜
background_removed_image = cv2.bitwise_and(rawimg_RT, rawimg_RT, mask=binary_mask)

# 显示去除背景后的图像
cv2.imshow('Background Removed Image', background_removed_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

这样修改后，类型不匹配的错误应该就会解决了。