def plotSmpAndSep(self):
self.lb5.setVisible(False)
p = PlotCanvas()
sep = []
QMessageBox.information(self, "tip", "开始绘制请稍后...")
for i in range(3, 30, 3):
Rm = self.Rms[:i, :] # 截取前i 个样本
Rs = self.Rss[:i, :]
pds = PDS(Rm, Rs) # 由选取的i 组样本进行DS 建模
win = self.win
pds.model(win, self.k)
Rsun = self.Rss[i:i * 2, :] # 选取子机上同样n 组未知样本进行DS 预测
Rmun = self.Rms[i:i * 2, :]
Rsunp = pds.predict(Rsun)
sep.append(p.computeSEP(Rmun[:, win:Rsun.shape[1] - win], Rsunp))
x = range(3, 30, 3)
p.axes.plot(x,
np.array(sep),
linewidth=3,
color='b',
marker='o',
markerfacecolor='red',
markersize=12)
p.axes.set_xlabel("样本数目（：组）")
p.axes.set_ylabel("平均预测标准误差（SEP）")
p.axes.set_title("样本数量 与 SEP 关系图")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
self.lb4.setText("绘制完成")
pass
def plotKAndSep(self):
self.lb5.setVisible(False)
p = PlotCanvas()
sep = []
QMessageBox.information(self, "tip", "开始绘制请稍后...")
# 样本数量定为当前样本数量
n = 10
Rm = self.Rms[:n, :] # 截取前n 个样本
Rs = self.Rss[:n, :]
pds = PDS(Rm, Rs) # 由选取的i 组样本进行DS 建模
win = self.win
Rsun = self.Rss[n:n * 2, :] # 选取子机上同样n 组未知样本进行PDS 预测
Rmswin = self.Rms[:, win:Rsun.shape[1] - win]
Rmun = Rmswin[n:n * 2, :]
for i in range(1, n + 1):
pds.model(win, i)
Rsunp = pds.predict(Rsun)
sep.append(p.computeSEP(Rmun, Rsunp))
x = range(1, n + 1)
p.axes.plot(x,
np.array(sep),
linewidth=3,
color='b',
marker='o',
markerfacecolor='red',
markersize=12)
p.axes.set_xlabel("主成分个数")
p.axes.set_ylabel("平均预测标准误差（SEP）")
p.axes.set_title("主成分个数 与 SEP 关系图")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
self.lb4.setText("绘制完成")
pass
def plotPDS(self):
pds = PDS(self.Rm, self.Rs)
# 窗口大小选择
pds.model(self.win, self.k) # pds 建模 得到pds.F
Rsun = self.Rss[:self.n, :] # 选取子机上同样n 组未知样本进行PDS 预测
Rsunp = pds.predict(Rsun) # Rsunp 为预测结果
win = self.win
# 舍弃窗口两端
Rmswin = self.Rms[:, win:Rsun.shape[1] - win] #Rmswin 源机上去掉两端的光谱
x = np.arange(1100, 2500, 2)
xwin = x[win:Rsun.shape[1] - win]
p = PlotCanvas()
# 差值
for i in range(Rsunp.shape[0]):
y1 = p.averSpec(Rmswin) # 源机均值光谱
p.axes.plot(xwin, y1, 'r-', label='源机均值光谱')
y2 = Rsunp[i, :] # 从子机预测的光谱
p.axes.plot(xwin, y2, label='样本' + str(i + 1) + '预测值')
# 差值
y = y2 - y1 # 目标机光谱
p.axes.plot(xwin, y, label='样本' + str(i + 1) + '预测偏差')
p.axes.legend()
p.axes.set_title("PDS算法预测结果")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
Rmun = Rmswin[:self.n, :]
sep = p.computeSEP(Rmun, Rsunp)
self.lb5.setVisible(True)
self.lb5.setText('%.5f' % sep)
pass
def plotWinAndSep(self):
self.lb5.setVisible(False)
p = PlotCanvas()
sep = []
QMessageBox.information(self, "tip", "开始绘制请稍后...")
# 样本数量定为当前样本数量
n = self.n
Rm = self.Rms[:n, :] # 截取前n 个样本
Rs = self.Rss[:n, :]
sks = Shenks(Rm, Rs) # 由选取的n 组样本进行sks 建模.
Rsun = self.Rss[n:n * 2, :] # 选取子机上同样n 组未知样本进行sks 预测
for i in range(1, 21):
win = i
sks.model(win)
Rsunp = sks.predict(Rsun)
xnew = sks.xnew
Rmswin = self.Rms[:, win:Rsun.shape[1] - win]
Rmswin = Rmswin[:, :xnew.shape[0]]
Rmun = Rmswin[n:n * 2, :]
sep.append(p.computeSEP(Rmun, Rsunp))
x = range(1, 21)
p.axes.plot(x,
np.array(sep),
linewidth=3,
color='b',
marker='o',
markerfacecolor='red',
markersize=12)
p.axes.set_xlabel("窗口大小")
p.axes.set_ylabel("平均预测标准误差（SEP）")
p.axes.set_title("窗口大小 与 SEP 关系图")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
self.lb4.setText("绘制完成")
pass
def plotSmpAndSep(self):
self.lb5.setVisible(False)
p = PlotCanvas()
sep = []
self.lb4.setText("绘制中......")
for i in range(3, 30, 3):
Rm = self.Rms[:i, :] # 截取前i 个样本
Rs = self.Rss[:i, :]
mlr = MLR(Rs, Rm) # 由选取的n 组样本进行DS 建模
mlr.modelling()
Rsun = self.Rss[i:i * 2, :] # 选取子机上同样n 组未知样本进行DS 预测
Rmun = self.Rms[i:i * 2, :]
Rsunp = np.dot(Rsun, mlr.A) # A 为得到的转换矩阵，Rsunp 为预测结果
sep.append(p.computeSEP(Rmun, Rsunp))
x = range(3, 30, 3)
p.axes.plot(x,
np.array(sep),
linewidth=3,
color='b',
marker='o',
markerfacecolor='red',
markersize=12)
p.axes.set_xlabel("样本数目（：组）")
p.axes.set_ylabel("平均预测标准误差（SEP）")
p.axes.set_title("样本数量 与 SEP 关系图")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
self.lb4.setText("绘制完成")
pass
def plotShenks(self):
win = self.win
sks = Shenks(self.Rm, self.Rs)
# 窗口大小选择
sks.model(win) # sks 建模
xnew = sks.xnew # 由一元二次模型校正后的波长
ynew = sks.ynew # 由校正后波长插值后的吸光度
Rsun = self.Rss[:self.n, :] # 选取子机上同样n 组未知样本进行Sks预测
Rsunp = sks.predict(Rsun)
x = np.arange(1100, 2500, 2)
# 舍弃窗口两端
xnew = sks.xnew # 由校正模型 校正后的 波长
Rmswin = self.Rms[:, win:Rsun.shape[1] - win]
p = PlotCanvas()
y1 = p.averSpec(Rmswin) # 源机均值光谱
y1 = y1[:xnew.shape[0]] # Rmswin 源机上去掉多余波长的光谱
# 差值
for i in range(Rsunp.shape[0]):
p.axes.plot(xnew, y1, 'r-', label='源机均值光谱')
y2 = Rsunp[i] # 从子机预测的光谱
p.axes.plot(xnew, y2, label='样本' + str(i + 1) + '预测值')
# 差值
y = y2 - y1 # 目标机光谱
p.axes.plot(xnew, y, label='样本' + str(i + 1) + '预测偏差')
p.axes.legend()
p.axes.set_title("Shenk's算法预测结果")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
Rmswin = Rmswin[:, :xnew.shape[0]]
Rmun = Rmswin[:self.n, :]
sep = p.computeSEP(Rmun, Rsunp)
self.lb5.setVisible(True)
self.lb5.setText('%.5f' % sep)
pass
def plotDS(self):
p = PlotCanvas()
mlr = MLR(self.Rs, self.Rm) # 由选取的n 组样本进行DS 建模
mlr.modelling()
Rsun = self.Rss[:self.n, :] # 选取子机上同样n 组未知样本进行DS 预测
Rsunp = np.dot(Rsun, mlr.A) # A 为得到的转换矩阵，Rsunp 为预测结果
x = np.arange(1100, 2500, 2)
for i in range(Rsunp.shape[0]):
y1 = p.averSpec(self.Rms) # 源机均值光谱
p.axes.plot(x, y1, 'r-', label='源机均值光谱')
y2 = Rsunp[i, :] # 从子机预测的光谱
p.axes.plot(x, y2, label='样本' + str(i + 1) + '预测值')
# 差值
y = y2 - y1 # 目标机光谱
p.axes.plot(x, y, label='样本' + str(i + 1) + '预测偏差')
p.axes.legend()
p.axes.set_title("DS算法预测结果")
self.hlayout4.replaceWidget(self.wg, p)
self.wg = p
Rmun = self.Rms[:self.n, :]
sep = p.computeSEP(Rmun, Rsunp)
self.lb5.setVisible(True)
self.lb5.setText('%.5f' % sep)
pass