potato-classification/functions.py at main · Kohmann/potato-classification · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
#!/usr/bin/env python
# coding: utf-8


import numpy as np
from spectral import principal_components
import matplotlib.pyplot as plt
import cv2
import spectral as sp


def plotSpectrum(arr, xaxis="wavelength", linelabel=0 ):

    if xaxis == "wavelength":
        x = np.linspace(400,1000,len(arr))
    elif xaxis == "band":
        x = np.linspace(0,len(arr), len(arr))

    else:
        print("xaxis wasnt wavelength or band")
        return 0

    if linelabel != 0:
        plt.plot(x, arr, label=linelabel)
        plt.legend()
    else:
        plt.plot(x, arr)
    plt.ylabel("Reflectance")
    plt.xlabel(xaxis)


def PCAreduce(img, frac=0.999):

    pc = principal_components(img)
    reduced = pc.reduce(fraction=frac)

    return reduced.transform(img), reduced.eigenvalues

def clip(img, mi, ma):
    """
    Clips values in between min and max
    Values bigger then max gets set equal to 0
    """
    for i in range(img.shape[2]):
        img[:,:,i][img[:,:,i] > ma] = 0
        img[:,:,i][img[:,:,i] < mi] = 0

    return 0

def meanSpectrum(cube):  # works only for subregions
    bands = cube.shape[2]
    avgSpec = np.zeros(bands)
    for i in range(bands):
        avgSpec[i] = cube[:,:,i].mean()
    return avgSpec

def getImage(time):
    files = [ "../Strawberry/Normalised/STR_B1_before-hit_VNIR_1800_SN00841_HSNR2_11998us_2020-08-02T184054_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_0mn_VNIR_1800_SN00841_HSNR2_11998us_2020-08-02T185621_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_30min_VNIR_1800_SN00841_HSNR2_11998us_2020-08-02T193019_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_1hr_VNIR_1800_SN00841_HSNR2_11998us_2020-08-02T202126_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_3hr_VNIR_1800_SN00841_HSNR2_11998us_2020-08-02T214153_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_12h_VNIR_1800_SN00841_HSNR2_9998us_2020-08-03T112003_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_24h_VNIR_1800_SN00841_HSNR2_9998us_2020-08-03T181621_raw_rad_nn.hdr",
             "../Strawberry/Normalised/STR_B1_48h_VNIR_1800_SN00841_HSNR2_9998us_2020-08-04T152605_raw_rad_nn.hdr"
            ]
    time = "_" + time # string must have _ in front to make it unique
    for file in files:
        if time in file:
            return file

    print('Format must be [before_hit, 0m, 30m, 1h, 3h, 12h, 24h, 48h]')
    return 0


def getRegion(img, x=0, y=0, kernel=0):
    if (not x or not y):
        print("Format is [image, x-coord, y-coord, kernel]")
    elif (x and y):
        return img[y - kernel//2 : y + kernel//2 + 1, x - kernel//2 : x + kernel//2 + 1]


def removeSeeds(img):

    if len(img.shape) == 3:
        img8bit = (img[:,:,img.shape[-1]-1]*255).astype('uint8')
    elif len(img.shape) == 2:
        img8bit = (img*255).astype('uint8')
    else:
        raise ValueError('Image is not 2d or 3d')

    thresh = cv2.threshold(img8bit,0,1, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1] # inverse otsu threshold to remove seeds
    erode = cv2.erode(thresh,np.ones((5,5),np.uint8)) # Removes the pixels that were next to the seeds

    return img[np.where(erode==1)]# returns pixel spectrums that are not seeds


def strawberryMask(img):
    imgBand = img[:,:,img.shape[-1]-1] # Image based on the last band
    img8bit = (imgBand*255).astype('uint8')  # Converting to 8bit
    tresh = cv2.threshold(img8bit,10,1, cv2.THRESH_BINARY)[1]  #
    closing  = cv2.morphologyEx(tresh, cv2.MORPH_CLOSE, np.ones((7,7),np.uint8), iterations=1)
    eroded = cv2.erode(closing, np.ones((7,7),np.uint8), iterations = 10)
    return eroded