谨慎使用效率极低的std::list.size()函数

import os import rasterio import numpy as np from scipy.stats import pearsonr from tqdm import tqdm # ---------------------------- 通用函数定义 ---------------------------- def preprocess_data(data): """数据预处理，将无效值替换为NaN""" invalid_value = -3.4028230607370965e+38 data[data == invalid_value] = np.nan return data def z_score_normalize(data): """Z-score标准化（含零标准差处理）""" mean = np.nanmean(data) std = np.nanstd(data) # 处理无效标准差 if std == 0 or np.isnan(std): normalized = np.zeros_like(data) else: normalized = (data - mean) / std # 保留原始NaN位置 normalized[np.isnan(data)] = np.nan return normalized def local_pearson(raster1, raster2, window_size=5): """计算局部Pearson相关系数（含窗口有效性检查）""" rows, cols = raster1.shape corr_matrix = np.full((rows, cols), np.nan) p_matrix = np.full((rows, cols), np.nan) half_win = window_size // 2 # 镜像填充处理边缘 raster1_padded = np.pad(raster1, half_win, mode='reflect') raster2_padded = np.pad(raster2, half_win, mode='reflect') for i in tqdm(range(rows), desc="计算进度"): for j in range(cols): # 提取窗口数据 win1 = raster1_padded[i:i + window_size, j:j + window_size].flatten() win2 = raster2_padded[i:i + window_size, j:j + window_size].flatten() # 剔除无效值 mask = ~np.isnan(win1) & ~np.isnan(win2) valid_win1 = win1[mask] valid_win2 = win2[mask] # 打印窗口数据信息 valid_count = len(valid_win1) std_win1 = np.std(valid_win1) if valid_count > 0 else np.nan std_win2 = np.std(valid_win2) if valid_count > 0 else np.nan print(f"窗口 ({i}, {j}): 有效数据点数量: {valid_count}, 标准差1: {std_win1}, 标准差2: {std_win2}") # 严格有效性检查 if len(valid_win1) < 3 or len(valid_win2) < 3: continue if np.std(valid_win1) == 0 or np.std(valid_win2) == 0: continue # 跳过零标准差窗口 # 计算相关系数和p值