使用trackpy和pims处理TIFF序列时遇OSError(Errno9)错误求助

问题 #

使用Python的trackpy和pims库追踪流体粒子位置，处理.tiff序列帧时，执行tp.batch出现OSError: [Errno 9] Bad file descriptor错误，且每次报错的帧不固定。

代码示例 #

import numpy as np
%matplotlib inline
import matplotlib as mpl
import matplotlib.pyplot as plt
from scipy.ndimage.interpolation import zoom
import pandas as pd
import trackpy as tp
import pims
from scipy.optimize import curve_fit

v = pims.ImageSequence('Capture_*.tiff')
v = pims.as_grey(v)

f = tp.locate(v[0], 5, minmass=20, maxsize=3, invert=True)

f = tp.batch(v[:], 5, minmass=20, maxsize=3, invert=True)

报错信息 #

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
/var/folders/xj/cw6n24v100l7s1rkm85p4qym0000gn/T/ipykernel_3149/1176320070.py in 
----> 1 f=tp.batch(v[:], 5, minmass=20, maxsize=3, invert=True)

~/opt/anaconda3/envs/APL_BM_env2/lib/python3.7/site-packages/trackpy/feature.py in batch(frames, diameter, output, meta, processes, after_locate, **kwargs)
    555     try:
    556         all_features = []
--> 557         for i, features in enumerate(map_func(curried_locate, frames)):
    558             image = frames[i]
    559             if hasattr(image, 'frame_no') and image.frame_no is not None:

~/opt/anaconda3/envs/APL_BM_env2/lib/python3.7/multiprocessing/pool.py in next(self, timeout)
    746         if success:
    747             return value
--> 748         raise value
    749 
    750     __next__ = next                    # XXX

OSError: [Errno 9] Bad file descriptor

解决思路与方案 #

该错误多由tp.batch默认多进程模式下，pims的ImageSequence对象出现文件句柄共享冲突导致，可尝试以下方案：

• 强制单进程运行：在tp.batch中指定processes=1，避免多进程间的文件资源竞争：

  f = tp.batch(v[:], 5, minmass=20, maxsize=3, invert=True, processes=1)
  

• 预先加载所有帧到内存：将所有tiff帧读取到内存列表后再传入tp.batch，绕开多进程重复读取文件的问题：

  frames_list = [frame for frame in v]
  f = tp.batch(frames_list, 5, minmass=20, maxsize=3, invert=True)
  

• 验证文件完整性：随机抽取报错帧单独读取，确认.tiff文件无损坏：

  # 替换为报错的帧序号
  test_frame = v[10]
  plt.imshow(test_frame)
  plt.show()
  

• 更新依赖库：旧版本trackpy/pims可能存在多进程兼容bug，更新至最新稳定版：

  pip install --upgrade trackpy pims
  

内容的提问来源于stack exchange，提问作者NikNack