import numpy as np
import matplotlib.pyplot as plt
import cv2
import pylab as mpl  # import matplotlib as mpl
from scipy.signal import savgol_filter
from signal_process.basic_signal_pro import normalize_fun
a = cv2.imread(r'.jpg')
a_red = a[:, :, 2]


def get_point_2(img):
    down = cv2.pyrDown(img)
    down = cv2.pyrDown(down)
    thresh, pic = cv2.threshold(down, 0, 255, cv2.THRESH_OTSU + cv2.THRESH_BINARY)
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
    erode = cv2.erode(pic, kernel)
    dilate = cv2.dilate(erode, kernel)
    cor = np.where(dilate > 0)
    print(cor)
    # 图像坐标系（x，-y）转换到矩阵坐标系(row, col)
    arr_col = -cor[0]
    arr_row = cor[1]
    # plt.scatter(arr_row, arr_col)
    # 没有问题，执行下一步。观察scatter图像，发现图像较粗的原因是一个x对应多个y，那么就需要做x的删减操作
    # 也就是使一个arr_row对应一个arr_col，具体选第一个出现的

    # step_1: 以x坐标值做排序
    arr_new = np.c_[arr_row, arr_col]
    arr_sort = arr_new[np.argsort(arr_new[:, 0])]
    # arr_sort = np.sort(arr_new)
    # 打印一下
    # print(arr_sort[:, 0][0:100])    # 结果符合预期

    # step_2:同一个x对应下的y只要一个
    down_x = []
    down_y = []
    for i in range(1, len(arr_row)):
        if arr_sort[i, 0] != arr_sort[i-1, 0]:
            down_x.append(arr_sort[i, 0])
            down_y.append(arr_sort[i, 1])

    # 画一下线,可以,但是线条不好看
    # plt.plot(down_x, down_y)

    # 做一个平滑处理
    sav_y = savgol_filter(down_y, 15, 5)
    return sav_y
    # plt.plot(sav_y)
    # plt.show()

原图片长这样，看看有没有搞得更好

我处理出来的