add: fit2 and fit3;finish frame work

cal.txt

@@ -0,0 +1,40 @@
+0 0 0.01006 0.01002 -1.29111
+0 1 0.00994 0.01009 -1.08340
+0 2 0.01050 0.01024 -1.15601
+0 3 0.00998 0.00984 -1.19307
+0 4 0.01023 0.01009 -0.97504
+0 5 0.00999 0.01011 -1.18887
+0 6 0.01008 0.00989 -0.90181
+0 7 0.00998 0.01002 -1.31894
+1 0 0.01009 0.00998 -0.84118
+1 1 0.01000 0.00988 -1.32354
+1 2 0.00998 0.00997 -1.41626
+1 3 0.00977 0.00994 -1.38833
+1 4 0.01003 0.01026 -1.35276
+1 5 0.00997 0.00981 -1.18387
+1 6 0.00995 0.00987 -1.02832
+1 7 0.01036 0.01005 -0.56795
+2 0 0.01019 0.01026 -0.09668
+2 1 0.01021 0.01024 -0.41860
+2 2 0.01033 0.01006 0.23775
+2 3 0.01027 0.01000 -0.47241
+2 4 0.01003 0.01032 -0.22038
+2 5 0.01030 0.01034 -0.39546
+2 6 0.01004 0.01014 -0.29457
+2 7 0.01089 0.01086 -0.92415
+3 0 0.01071 0.01037 -0.46705
+3 1 0.01092 0.01074 -1.06885
+3 2 0.01053 0.01067 -0.68895
+3 3 0.01059 0.01080 -0.96538
+3 4 0.01098 0.01058 -0.51257
+3 5 0.01117 0.01110 -1.20546
+3 6 0.01042 0.01085 -1.06498
+3 7 0.01055 0.01023 -1.16872
+4 0 0.01017 0.01043 -0.73180
+4 1 0.01034 0.01005 -0.89867
+4 2 0.01039 0.01029 -0.59282
+4 3 0.01038 0.01029 -0.83104
+4 4 0.01019 0.01043 -0.56346
+4 5 0.01065 0.01008 -1.12639
+4 6 0.01041 0.01034 -0.64923
+4 7 0.01000 0.01019 -1.03055

main.py

@@ -1,33 +1,79 @@
 import os
 import re
-import numpy as np
+from qdx import Bind
+from qdx.utils import get_hist, file_filter
 from tqdm import tqdm
-from qdx import BindFilter, Bind
+from matplotlib import pyplot as plt
-
-binds = []
-BF = BindFilter()
 
 path = 'result/bind'
 file_list = os.listdir(path)
-
 reg = re.compile(r'(([0-9]{4})-([0-9])-([0-9])).txt')
 
-pbar = tqdm(desc="Gaussian Mixture Bind Filter", total=len(file_list))
+n, m = 5, 8
+binds = [[j for j in range(m)] for i in range(n)]
+for i in range(n):
+    for j in range(m):
+        binds[i][j] = Bind(i, j)
+
+
+pbar = tqdm(desc="Gaussian Mixture Filter", total=len(file_list))
 for file in file_list:
-    name, E, n, m = reg.match(file).groups()
+    bind = binds[int(i)][int(j)]
+
+    name, E, i, j = reg.match(file).groups()
     file = os.path.join(path, file)
 
-    BF.filter(file)
+    _, data = file_filter(file)
-    # BF.draw('result/GMM/' + name + '.png')
+    # bind.draw_filter('result/GMM/' + name + '.png')
-    # np.savetxt('result/bind-GMM/' + name + '.txt', BF.fit_data, fmt='%d')
+    # np.savetxt('result/bind-GMM/' + name + '.txt', data, fmt='%d')
-    binds.append(Bind(int(E), int(n), int(m), BF.fit_data))
+    bind.add_data(int(E) / 100, data)
 
     pbar.update(1)
 pbar.close()
 
-pbar = tqdm(desc="Bind Linear Fit", total=len(binds))
+
-for bind in binds:
+split_n = [8, 7, 7, 8, 8]
-    bind.fit()
+pbar = tqdm(desc="Bind Process", total=n * m)
-    bind.draw('result/L-FIT/' + bind.name + '.png')
+for i in range(n):
+    for j in range(m):
+        bind: Bind = binds[i][j]
+        bind.filter()
+        bind.fit1()
+        bind.solve()
+        bind.split(n=split_n[i])
+        bind.fit2()
+        bind.draw_fit1('result/FIT1/' + bind.name + '.png')
+        bind.draw_fit2('result/FIT2/' + bind.name + '.png')
         pbar.update(1)
 pbar.close()
+
+
+y = [
+    [620, 610, 600, 590, 580, 570, 560, 550],
+    [485, 475, 465, 455, 445, 435, 425],
+    [360, 350, 340, 330, 320, 310, 300],
+    [235, 225, 215, 205, 195, 185, 175, 165],
+    [100,  90,  80,  70,  60,  50,  40,  30]
+]
+for i in range(n):
+    k = 1
+    fig = plt.figure(figsize=(14, 6))
+    for j in range(m):
+        ax = fig.add_subplot(int(m / 2), 2, k)
+
+        bind = binds[i][j]
+        bind.fit3(y[i])
+        ny = bind.reg3.predict(bind.x1)
+        c, e = get_hist(ny, delta=0.5)
+
+        ax.scatter(e, c, s=0.5)
+        k += 1
+    fig.savefig('./Figure-' + str(i) + '.png', facecolor='w', transparent=False)
+    plt.tight_layout()
+    plt.close()
+
+f = open('cal.txt', 'w')
+for i in range(n):
+    for j in range(m):
+        bind = binds[i][j]
+        f.writelines('{:d} {:d} {:.5f} {:.5f} {:.5f}\n'.format(i, j, bind.k1, bind.k2, bind.b))

qdx/Bind.py

@@ -3,16 +3,7 @@ import matplotlib.pyplot as plt
 from astropy.modeling import models, fitting
 from sklearn.mixture import GaussianMixture
 from sklearn.linear_model import LinearRegression
-
+from .utils import get_hist
-def get_hist(data, maxN=50):
-    step = 1
-    edge = np.arange(data.min(), data.max() + 1, step)
-    count, _ = np.histogram(data, bins=edge)
-    while count.max() <= maxN:
-        step += 1
-        edge = np.arange(data.min(), data.max() + 1, step)
-        count, _ = np.histogram(data, bins=edge)
-    return count, (edge[1:] + edge[:-1]) / 2
 
 class Bind(object):
     flag = 0
@@ -32,7 +23,7 @@ class Bind(object):
             self.E2 = E
             self.x2, self.y2 = data[:, 0].reshape(-1, 1), data[:, 1].reshape(-1, 1)
 
-    def fit(self):
+    def fit1(self):
         self.reg1 = LinearRegression()
         self.reg1.fit(self.x1, self.y1)
         self.K1 = self.reg1.coef_[0][0]
@@ -43,13 +34,26 @@ class Bind(object):
         self.K2 = self.reg2.coef_[0][0]
         self.C2 = self.reg2.intercept_[0]
 
+    def filter(self):
+        self.fit1()
+        ny1 = self.y1 - self.reg1.predict(self.x1)
+        ny2 = self.y2 - self.reg2.predict(self.x2)
+
+        idx1 = np.where(ny1 <= ny1.std())[0]
+        idx2 = np.where(ny2 <= ny2.std())[0]
+
+        self.x1 = self.x1[idx1]
+        self.y1 = self.y1[idx1]
+        self.x2 = self.x2[idx2]
+        self.y2 = self.y2[idx2]
+
     def solve(self):
         self.k1 = (self.E1 - self.E2) / (self.C1 - self.C2)
-        self.k2 = -(self.K1 + self.K2) / (self.k1 * 2)
+        self.k2 = -(self.K1 + self.K2) * self.k1 / 2
         self.b = (self.E1 - self.C1 * self.k1 + self.E2 - self.C2 * self.k1) / 2
 
     def split(self, n=7):
-        self.cluster = []
+        self.clusters = []
         data = np.concatenate((self.x1, self.y1), axis=1)
 
         model = GaussianMixture(n_components=n)
@@ -58,16 +62,17 @@ class Bind(object):
         ny = model.predict(data)
         for i in np.unique(ny):
             idx = np.where(ny == i)[0]
-            self.cluster.append(data[idx])
+            self.clusters.append(data[idx])
 
-        self.cluster = np.array(self.cluster)
+        self.clusters = np.array(self.clusters, dtype=object)
+        self.clusters = sorted(self.clusters, key=lambda x: len(x))
 
     def fit2(self):
         self.fx = []
         self.fy = []
         fitter = fitting.LevMarLSQFitter()
 
-        for data in self.cluster:
+        for data in self.clusters:
             x1, x2 = data[:, 0], data[:, 1]
             c1, e1 = get_hist(x1)
             c2, e2 = get_hist(x2)
@@ -78,12 +83,16 @@ class Bind(object):
             self.fx.append(fitted_model1.mean.value)
             self.fy.append(fitted_model2.mean.value)
 
-    def pX(self, x1=None, x2=None):
+        idx = np.argsort(self.fx)
-        x1 = x1 if x1 else self.x1
-        x2 = x2 if x2 else self.y1
-        self.px = (self.k2 * x2 - self.k1 * x1) * self.L / self.E1 + self.C
 
-    def draw(self, title):
+    def fit3(self, y):
+        self.fx = np.sort(self.fx)[::-1].reshape(-1, 1)
+        self.reg3 = LinearRegression()
+        self.reg3.fit(self.fx, y)
+        self.K3 = self.reg3.coef_[0]
+        self.C3 = self.reg3.intercept_
+
+    def draw_fit1(self, title):
         fig = plt.figure(figsize=(8, 8))
         ax = fig.add_subplot(1, 1, 1)
         ax.scatter(self.x1, self.y1, s=0.1, c='black', label=r'$E_1$')
@@ -94,9 +103,32 @@ class Bind(object):
         fig.savefig(title, facecolor='w', transparent=False)
         plt.close()
 
+    def draw_split(self, title):
+        fig = plt.figure(figsize=(8, 8))
+        ax = fig.add_subplot(1, 1, 1)
+        for data in self.clusters:
+            ax.scatter(data[:, 0], data[:, 1], s=0.1)
+        fig.savefig(title, facecolor='w', transparent=False)
+        plt.close()
+
+    def draw_fit2(self, title):
+        k, n = 1, len(self.clusters)
+        fig = plt.figure(figsize=(8, 10))
+        for data in self.clusters:
+            ax = fig.add_subplot(n, 1, k)
+            x1, x2 = data[:, 0], data[:, 1]
+            c1, e1 = get_hist(x1)
+            c2, e2 = get_hist(x2)
+            ax.scatter(e1, c1, s=0.1)
+            ax.scatter(e2, c2, s=0.1)
+            k += 1
+        plt.tight_layout()
+        fig.savefig(title, facecolor='w', transparent=False)
+        plt.close()
+
     @property
     def name(self):
-        return '{:d}-{:d}-{:d}-{:d}'.format(self.E1, self.E2, self.n, self.m)
+        return '{:.1f}-{:.1f}-{:d}-{:d}'.format(self.E1, self.E2, self.n, self.m)
 
     @property
     def RSquare1(self):

qdx/BindFilter.py

@@ -1,33 +0,0 @@
-import numpy as np
-from matplotlib import pyplot as plt
-from sklearn.mixture import GaussianMixture
-
-class BindFilter(object):
-    def __init__(self):
-        pass
-
-    def filter(self, file):
-        self.clusters = []
-        self.fit_data = np.array([])
-
-        data = np.loadtxt(file, dtype=np.uint16)
-
-        model = GaussianMixture(n_components=2)
-        model.fit(data)
-
-        ny = model.predict(data)
-        for i in np.unique(ny):
-            idx = np.where(ny == i)[0]
-            self.fit_data = idx if len(idx) > len(self.fit_data) else self.fit_data
-            self.clusters.append(data[idx])
-
-        self.data = data
-        self.fit_data = data[self.fit_data]
-
-    def draw(self, title):
-        fig = plt.figure(figsize=(8, 8))
-        ax = fig.add_subplot(1, 1, 1)
-        for cluster in self.clusters:
-            ax.scatter(cluster[:, 0], cluster[:, 1], s=0.1)
-        fig.savefig(title)
-        plt.close()

qdx/__init__.py

@@ -1,2 +1 @@
 from .Bind import Bind
-from .BindFilter import BindFilter

qdx/utils.py

@@ -0,0 +1,38 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.mixture import GaussianMixture
+
+def file_filter(file):
+    filter_data = []
+    fit_data = np.array([])
+
+    data = np.loadtxt(file, dtype=np.uint16)
+
+    model = GaussianMixture(n_components=2)
+    model.fit(data)
+
+    ny = model.predict(data)
+    for i in np.unique(ny):
+        idx = np.where(ny == i)[0]
+        fit_data = idx if len(idx) > len(fit_data) else fit_data
+        filter_data.append(data[idx])
+
+    return filter_data, data[fit_data]
+
+def draw_filter(data, title):
+    fig = plt.figure(figsize=(8, 8))
+    ax = fig.add_subplot(1, 1, 1)
+    for cluster in data:
+        ax.scatter(cluster[:, 0], cluster[:, 1], s=0.1)
+    fig.savefig(title, facecolor='w', transparent=False)
+    plt.close()
+
+def get_hist(data, delta=1, maxN=50):
+    step = delta
+    edge = np.arange(data.min(), data.max() + 1, step)
+    count, _ = np.histogram(data, bins=edge)
+    while count.max() <= maxN:
+        step += delta
+        edge = np.arange(data.min(), data.max() + 1, step)
+        count, _ = np.histogram(data, bins=edge)
+    return count, (edge[1:] + edge[:-1]) / 2