就快完成了，但是你的逻辑发送代码在双击时画一条线，而没有存储双击的位置，所以它需要两次单击来画一条线。另外，你需要在圆圈代码中画画布。下面是一个最小的修改版本，它完成了要求1和2：

import matplotlib.pyplot as plt

class LineDrawer(object):
    lines = []
    def draw_line(self, startx,starty):
        ax = plt.gca()
        xy = plt.ginput(1)
        x = [startx,xy[0][0]]
        y = [starty,xy[0][1]]
        line = plt.plot(x,y)
        ax.figure.canvas.draw()

        self.lines.append(line)

def onclick(event):
    if event.dblclick:
        if event.button == 1:
            # Draw line
            ld = LineDrawer()
            ld.draw_line(event.xdata,event.ydata) # here you click on the plot
        elif event.button == 3:
            # Write to figure
            plt.figtext(3, 8, 'boxed italics text in data coords', style='italic', bbox={'facecolor':'red', 'alpha':0.5, 'pad':10})
            circ = plt.Circle((event.xdata, event.ydata), radius=0.07, color='g')
            ax.add_patch(circ)
            ax.figure.canvas.draw()
        else:
            pass # Do nothing

def onpick(event):
    thisline = event.artist
    xdata = thisline.get_xdata()
    ydata = thisline.get_ydata()
    ind = event.ind
    print ('onpick points:', zip(xdata[ind], ydata[ind]))


fig, ax = plt.subplots()

connection_id = fig.canvas.mpl_connect('button_press_event', onclick)
fig.canvas.mpl_connect('pick_event', onpick)

plt.tight_layout()

plt.show()

注意matplotlib可能不是实现这些要求的最好或最简单的方法--而且轴在绘制第一条直线时会自动重新缩放。您可以通过修改xlim和ylim来改变这一点。例如：

ax.set_xlim([0,2])
ax.set_ylim([0,2])

为了实现要求3，你将不得不存储拾取的对象，并监听匹配delete的按键来删除它。下面是一个结合了以上所有内容的版本。我已经尽可能地坚持你的设计。我将对拾取的对象的引用存储在相关的axis对象中。如果你不想实现自己的数据结构来存储拾取的对象，你可能会想实现自己的数据结构。我不喜欢将它插入到当前轴中。我已经测试了一下，但是可能有一些点击/按键序列会混淆逻辑。

import matplotlib.pyplot as plt

# function to draw lines - from matplotlib examples.  Note you don't need
# to keep a reference to the lines drawn, so I've removed the class as it
# is overkill for your purposes
def draw_line(startx,starty):
        ax = plt.gca()
        xy = plt.ginput(1)
        x = [startx,xy[0][0]]
        y = [starty,xy[0][1]]
        line = ax.plot(x,y, picker=5) # note that picker=5 means a click within 5 pixels will "pick" the Line2D object
        ax.figure.canvas.draw()        

def onclick(event):
    """
    This implements click functionality.  If it's a double click do something,
    else ignore.
    Once in the double click block, if its a left click, wait for a further 
    click and draw a line between the double click co-ordinates and that click
    (using ginput(1) - the 1 means wait for one mouse input - a higher number
    is used to get multiple clicks to define a polyline)
    If the double click was a right click, draw the fixed radius circle

    """
    if event.dblclick:
        if event.button == 1:
            # Draw line
            draw_line(event.xdata,event.ydata) # here you click on the plot
        elif event.button == 3:
            # Write to figure
            plt.figtext(3, 8, 'boxed italics text in data coords', style='italic', bbox={'facecolor':'red', 'alpha':0.5, 'pad':10})
            circ = plt.Circle((event.xdata, event.ydata), radius=0.07, color='g', picker = True)
            ax.add_patch(circ)
            ax.figure.canvas.draw()
        else:
            pass # Do nothing

def onpick(event):    
    """
    Handles the pick event - if an object has been picked, store a
    reference to it.  We do this by simply adding a reference to it
    named 'stored_pick' to the axes object.  Note that in python we
    can dynamically add an attribute variable (stored_pick) to an 
    existing object - even one that is produced by a library as in this
    case
    """
    this_artist = event.artist #the picked object is available as event.artist
    # print(this_artist) #For debug just to show you which object is picked
    plt.gca().picked_object = this_artist

def on_key(event):
    """
    Function to be bound to the key press event
    If the key pressed is delete and there is a picked object,
    remove that object from the canvas
    """
    if event.key == u'delete':
        ax = plt.gca()
        if ax.picked_object:
            ax.picked_object.remove()
            ax.picked_object = None
            ax.figure.canvas.draw()

fig, ax = plt.subplots()

#First we need to catch three types of event, clicks, "picks" (a specialised
#type of click to select an object on a matplotlib canvas) and key presses.
#The logic is - if it's a right double click, wait for the next click and draw
#a line, if its a right double click draw a fixed radius circle.  If it's a
#pick, store a reference to the picked item until the next keypress.  If it's
#a keypress - test if it's delete and if so, remove the picked object.
#The functions (defined above) bound to the events implement this logic
connection_id = fig.canvas.mpl_connect('button_press_event', onclick)
fig.canvas.mpl_connect('pick_event', onpick)
cid = fig.canvas.mpl_connect('key_press_event', on_key)

#set the size of the matplotlib figure in data units, so that it doesn't
#auto-resize (which it will be default on the first drawn item)
ax.set_xlim([0,2])
ax.set_ylim([0,2])
ax.aspect = 1
plt.tight_layout()

plt.show()

我最近在做一个名为mpltoolbox的项目，它的目标是让在Matplotlib图形上绘制直线、点、矩形和椭圆变得更容易。也许这能让你在这里的一些事情变得更容易？
https://github.com/scipp/mpltoolbox