diff --git a/plugins/ai_method/basic_cd.py b/plugins/ai_method/basic_cd.py
index d772c85..07049af 100644
--- a/plugins/ai_method/basic_cd.py
+++ b/plugins/ai_method/basic_cd.py
@@ -125,16 +125,24 @@ class BasicAICD(AlgFrontend):
             block_size = (xsize, cell_size[1])
             if block_xy[1] + block_size[1] > ysize:
                 block_size = (xsize, ysize - block_xy[1])
+            
+            np.seterr(divide='ignore',invalid='ignore')
             block_data = temp_in_ds.ReadAsArray(*block_xy, *block_size)
+            max_diff=block_data.max()
+            min_diff=block_data.min()
+
             block_data = (block_data - min_diff) / (max_diff - min_diff) * 255
+            
             block_data = block_data.astype(np.uint8)
             out_normal_ds.GetRasterBand(1).WriteArray(block_data, *block_xy)
+            # os.system('pause')
             # hist_t, _ = np.histogram(block_data, bins=256, range=(0, 256))
             # hist += hist_t
         # print(hist)
         del temp_in_ds
         del out_normal_ds
         try:
+            # os.system('pause')
             os.remove(out_tif)
         except:
             pass
diff --git a/plugins/evaluation/main.py b/plugins/evaluation/main.py
index 1a772b1..3193e41 100644
--- a/plugins/evaluation/main.py
+++ b/plugins/evaluation/main.py
@@ -144,8 +144,9 @@ class EvaluationPlugin(BasicPlugin):
 
             for k in range(2):
                 for l in range(2):
-                    cfm[k,l] += np.sum((pred_block == k) & (gt_block >= l))
-        
+                    # cfm[k,l] += np.sum((pred_block == k) & (gt_block >= l))
+                    cfm[k,l] += np.sum((pred_block == k) & (gt_block == l))
+
         result_path = os.path.join(Project().other_path, f'{layer.name}_{os.path.basename(gt)}_evaluation.txt')
         with open(result_path, 'w', encoding='utf-8') as f:
             f.write(f'预测结果：{layer.path}\n')
@@ -169,7 +170,7 @@ class EvaluationPlugin(BasicPlugin):
             Kappa: {kappa(cfm)}
             ''')
             f.flush()
-            
+        # os.system('pause')    
         txt=f'''
             预测结果：{layer.path}\n
 
@@ -193,12 +194,13 @@ class EvaluationPlugin(BasicPlugin):
             F1:  {cfm[1,1] / (np.sum(cfm[1,:]) + np.sum(cfm[:,1]) - cfm[1,1])}
             Kappa: {kappa(cfm)}
             '''
+
+        # self.msgbox = Win(txt)
+        # self.msgbox.show()
         
-        self.msgbox = Win(txt)
-        self.msgbox.show()
         # os.system(f'c:/windows/notepad.exe "{result_path}"')
 
-        # self.send_message.emit('精度评估完成')
+        self.send_message.emit('精度评估完成,结果保存在'+result_path)
     
     def show_dialog(self):
         dialog = EvalutationDialog(self.mainwindow)
@@ -212,10 +214,13 @@ class EvaluationPlugin(BasicPlugin):
             t.start()
         
 from .window import Ui_Form
-class Win(QWidget,Ui_Form):
+# class Win(QWidget,Ui_Form):
+
+#     def __init__(self,txt):
+#         super().__init__()
+
+#         self.setupUi(self)
+#         self.label.setText(txt)
+
 
-    def __init__(self,txt):
-        super().__init__()
 
-        self.setupUi(self)
-        self.label.setText(txt)
\ No newline at end of file
diff --git a/plugins/export_to/main.py b/plugins/export_to/main.py
index 13637d6..e6b02eb 100644
--- a/plugins/export_to/main.py
+++ b/plugins/export_to/main.py
@@ -138,7 +138,7 @@ class ExportPlugin(BasicPlugin):
             [1, 255, 0, 0]
         ]
 
-        if self.style_path is '':
+        if self.style_path == '':
             style = default_style
         else:
             style = np.loadtxt(self.style_path, comments='#', delimiter=',')
@@ -160,27 +160,42 @@ class ExportPlugin(BasicPlugin):
         # TODO: check data range is 0-1
 
         def get_color(v):
-            first_color = []
-            second_color = []
-            first_value = 0
-            second_value = 1
-            for s in style:
-                if s[0] <= v:
-                    first_value = s[0]
-                    first_color = s[1:]
-                else:
-                    second_value = s[0]
-                    second_color = s[1:]
-                    break
-            if second_value == 1:
-                return np.array(style[-1][1:])
+            # first_color = []
+            # second_color = []
+            # first_value = 0
+            # second_value = 1
+            # for s in style:
+            #     if s[0] <= v:
+            #         first_value = s[0]
+            #         first_color = s[1:]
+            #     else:
+            #         second_value = s[0]
+            #         second_color = s[1:]
+            #         break
+            # if second_value == 1:
+            #     return np.array(style[-1][1:])
+
+            # first_dis = (v - first_value) / (second_value - first_value)
+            # second_dis = (second_value - v) /  (second_value - first_value)
+            # first_color = np.array(first_color)
+            # second_color = np.array(second_color)
+            # color = first_color* first_dis + second_color * second_dis
+            # return np.floor(color)
+        
+            if v==0:
+                return np.array(style[-1][1:])
+            elif v==1:
+                return np.array(style[-2][1:])
+            else:
+                first_color =style[-1][1:]
+                second_color=style[-2][1:]
+
+                first_color = np.array(first_color)
+                second_color=np.array(second_color)
+                color_inter=first_color*(1-v)+second_color*v
+                return color_inter
+
 
-            first_dis = (v - first_value) / (second_value - first_value)
-            second_dis = (second_value - v) /  (second_value - first_value)
-            first_color = np.array(first_color)
-            second_color = np.array(second_color)
-            color = first_color* first_dis + second_color * second_dis
-            return np.floor(color)
     
 
         get_color = np.frompyfunc(get_color, nin=1, nout=1)
diff --git a/plugins/filter_collection/lms_filter.py b/plugins/filter_collection/lms_filter.py
index 704ed77..777eef7 100644
--- a/plugins/filter_collection/lms_filter.py
+++ b/plugins/filter_collection/lms_filter.py
@@ -13,46 +13,37 @@ from datetime import datetime
 import numpy as np
 import torch
 import torch.nn.functional as F
-def adaptiveMedianDeNoise(count, original):
-    # 初始窗口大小
-    startWindow = 3
-    # 卷积范围
-    c = count // 2
-    rows, cols = original.shape
-    newI = np.zeros(original.shape)
-    # median = 
-
-    for i in range(c, rows - c):
-        for j in range(c, cols - c):
-            startWindow = 3 
-            k = int(startWindow / 2)
-            median = np.median(original[i - k:i + k + 1, j - k:j + k + 1])
-            mi = np.min(original[i - k:i + k + 1, j - k:j + k + 1])
-            ma = np.max(original[i - k:i + k + 1, j - k:j + k + 1])
-            if mi < median < ma:
-                if mi < original[i, j] < ma:
-                    newI[i, j] = original[i, j]
-                else:
-                    newI[i, j] = median
-
-            else:
-                while True:
-                    startWindow = startWindow + 2
-                    k = int(startWindow / 2)
-                    median = np.median(original[i - k:i + k + 1, j - k:j + k + 1])
-                    mi = np.min(original[i - k:i + k + 1, j - k:j + k + 1])
-                    ma = np.max(original[i - k:i + k + 1, j - k:j + k + 1])
-
-                    if mi < median < ma or startWindow >= count:
-                        break
-
-                if mi < median < ma or startWindow > count:
-                    if mi < original[i, j] < ma:
-                        newI[i, j] = original[i, j]
+import cv2
+from tqdm import tqdm
+def adaptiveMedianDeNoise(count, image):
+    origen = 3                                                        # 初始窗口大小
+    board = origen//2                                                 # 初始应扩充的边界
+    # max_board = max_size//2                                         # 最大可扩充的边界
+    copy = cv2.copyMakeBorder(image, *[board]*4, borderType=cv2.BORDER_DEFAULT)         # 扩充边界
+    out_img = np.zeros(image.shape)
+    for i in tqdm(range(image.shape[0])):
+        for j in range(image.shape[1]):
+            def sub_func(src, size):                         # 两个层次的子函数
+                kernel = src[i:i+size, j:j+size]
+                # print(kernel)
+                z_med = np.median(kernel)
+                z_max = np.max(kernel)
+                z_min = np.min(kernel)
+                if z_min < z_med < z_max:                                 # 层次A
+                    if z_min < image[i][j] < z_max:                       # 层次B
+                        return image[i][j]
                     else:
-                        newI[i, j] = median
+                        return z_med
+                else:
+                    next_size = cv2.copyMakeBorder(src, *[1]*4, borderType=cv2.BORDER_DEFAULT)   # 增尺寸
+                    size = size+2                                        # 奇数的核找中值才准确
+                    if size <= count:
+                        return sub_func(next_size, size)     # 重复层次A
+                    else:
+                        return z_med
+            out_img[i][j] = sub_func(copy, origen)
 
-    return newI
+    return out_img
 
 @FILTER.register
 class AdaptiveFilter(AlgFrontend):
diff --git a/plugins/filter_collection/lmsnp_filter.py b/plugins/filter_collection/lmsnp_filter.py
index 8d773d8..8927bbe 100644
--- a/plugins/filter_collection/lmsnp_filter.py
+++ b/plugins/filter_collection/lmsnp_filter.py
@@ -13,45 +13,38 @@ from datetime import datetime
 import numpy as np
 import torch
 import torch.nn.functional as F
-def adaptiveMedianDeNoise(count, original):
-    # 初始窗口大小
-    startWindow = 3
-    # 卷积范围
-    c = count // 2
-    rows, cols = original.shape
-    newI = np.zeros(original.shape)
-    # median = 
-
-    for i in range(c, rows - c):
-        for j in range(c, cols - c):
-            k = int(startWindow / 2)
-            median = np.median(original[i - k:i + k + 1, j - k:j + k + 1])
-            mi = np.min(original[i - k:i + k + 1, j - k:j + k + 1])
-            ma = np.max(original[i - k:i + k + 1, j - k:j + k + 1])
-            if mi < median < ma:
-                if mi < original[i, j] < ma:
-                    newI[i, j] = original[i, j]
-                else:
-                    newI[i, j] = median
-
-            else:
-                while True:
-                    startWindow = startWindow + 2
-                    k = int(startWindow / 2)
-                    median = np.median(original[i - k:i + k + 1, j - k:j + k + 1])
-                    mi = np.min(original[i - k:i + k + 1, j - k:j + k + 1])
-                    ma = np.max(original[i - k:i + k + 1, j - k:j + k + 1])
-
-                    if mi < median < ma or startWindow >= count:
-                        break
-
-                if mi < median < ma or startWindow >= count:
-                    if mi < original[i, j] < ma:
-                        newI[i, j] = original[i, j]
+import pandas as pd
+import cv2
+from tqdm import tqdm
+def adaptiveMedianDeNoise(count, image):
+    origen = 3                                                        # 初始窗口大小
+    board = origen//2                                                 # 初始应扩充的边界
+    # max_board = max_size//2                                         # 最大可扩充的边界
+    copy = cv2.copyMakeBorder(image, *[board]*4, borderType=cv2.BORDER_DEFAULT)         # 扩充边界
+    out_img = np.zeros(image.shape)
+    for i in tqdm(range(image.shape[0])):
+        for j in range(image.shape[1]):
+            def sub_func(src, size):                         # 两个层次的子函数
+                kernel = src[i:i+size, j:j+size]
+                # print(kernel)
+                z_med = np.median(kernel)
+                z_max = np.max(kernel)
+                z_min = np.min(kernel)
+                if z_min < z_med < z_max:                                 # 层次A
+                    if z_min < image[i][j] < z_max:                       # 层次B
+                        return image[i][j]
                     else:
-                        newI[i, j] = median
+                        return z_med
+                else:
+                    next_size = cv2.copyMakeBorder(src, *[1]*4, borderType=cv2.BORDER_DEFAULT)   # 增尺寸
+                    size = size+2                                        # 奇数的核找中值才准确
+                    if size <= count:
+                        return sub_func(next_size, size)     # 重复层次A
+                    else:
+                        return z_med
+            out_img[i][j] = sub_func(copy, origen)
 
-    return newI
+    return out_img
 
 @FILTER.register
 class AdaptiveNPFilter(AlgFrontend):
diff --git a/plugins/thres/main.py b/plugins/thres/main.py
index 3dc7d73..e1ea94b 100644
--- a/plugins/thres/main.py
+++ b/plugins/thres/main.py
@@ -138,7 +138,7 @@ class OTSUAlg(AlgFrontend):
         band = ds.GetRasterBand(1)
         # band_count = ds.RasterCount
         
-        hist = np.zeros(256, dtype=np.int)
+        hist = np.zeros(256, dtype=np.int_)
         xsize = ds.RasterXSize
         ysize = ds.RasterYSize