122 lines
4.4 KiB
Python
122 lines
4.4 KiB
Python
import cv2
|
|
import numpy as np
|
|
import joblib
|
|
from skimage.measure import regionprops, label
|
|
|
|
# Fungsi untuk menghitung eccentricity
|
|
def calculate_eccentricity(region):
|
|
if region.minor_axis_length == 0:
|
|
return 0
|
|
else:
|
|
return np.sqrt(1 - (region.minor_axis_length / region.major_axis_length)**2)
|
|
|
|
# Load model dan LabelEncoder yang telah disimpan
|
|
model_path = 'D:/Skripsi/ekstraksi fitur/rf_model.joblib'
|
|
label_encoder_path = 'D:/Skripsi/ekstraksi fitur/label_encoder.joblib'
|
|
rf_model = joblib.load(model_path)
|
|
le = joblib.load(label_encoder_path)
|
|
|
|
# Fungsi untuk mengekstraksi fitur dari gambar tunggal dengan preprocessing khusus
|
|
def extract_features(image_path, label_name):
|
|
img = cv2.imread(image_path, 1)
|
|
if img is None:
|
|
print(f"Gagal membaca file: {image_path}")
|
|
return None
|
|
|
|
blue, green, red = cv2.split(img)
|
|
|
|
# Thresholding dan preprocessing khusus untuk setiap jenis bakteri
|
|
if label_name == 'Corynebacterium Diphteriae':
|
|
ret, img_bin = cv2.threshold(green, 115, 255, cv2.THRESH_BINARY_INV)
|
|
img_bin = cv2.erode(cv2.dilate(img_bin, None, iterations=1), None, iterations=4)
|
|
elif label_name == 'Mycobacterium Tuberculosis':
|
|
ret, img_bin = cv2.threshold(green, 135, 255, cv2.THRESH_BINARY_INV)
|
|
img_bin = cv2.dilate(img_bin, None, iterations=1)
|
|
|
|
num_labels, labels, stats, centroids = cv2.connectedComponentsWithStats(img_bin, connectivity=8)
|
|
mask = np.zeros_like(img_bin)
|
|
min_size = 50
|
|
max_size = 400
|
|
|
|
for j in range(1, num_labels):
|
|
if min_size <= stats[j, cv2.CC_STAT_AREA] <= max_size:
|
|
mask[labels == j] = 255
|
|
|
|
img_bin = mask
|
|
elif label_name == 'Neiseria Gonorroea':
|
|
ret, img_bin = cv2.threshold(green, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
|
|
img_bin = cv2.dilate(img_bin, None, iterations=1)
|
|
|
|
num_labels, labels, stats, centroids = cv2.connectedComponentsWithStats(img_bin, connectivity=8)
|
|
mask = np.zeros_like(img_bin)
|
|
min_size = 200
|
|
max_size = 2500
|
|
|
|
for j in range(1, num_labels):
|
|
if min_size <= stats[j, cv2.CC_STAT_AREA] <= max_size:
|
|
mask[labels == j] = 255
|
|
|
|
img_bin = mask
|
|
elif label_name == 'Staphylococcus Aureus':
|
|
ret, img1 = cv2.threshold(green, 110, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
|
|
img_bin = cv2.erode(cv2.dilate(img1.copy(), None, iterations=1), None, iterations=1)
|
|
img_bin = cv2.morphologyEx(img_bin.copy(), cv2.MORPH_OPEN, None)
|
|
elif label_name == 'Streptococcus Pneumoniae':
|
|
ret, img1 = cv2.threshold(green, 110, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
|
|
img_bin = cv2.erode(img1.copy(), None, iterations=1)
|
|
img_bin = cv2.dilate(img1.copy(), None, iterations=1)
|
|
|
|
# Labeling
|
|
labeled_img = label(img_bin)
|
|
regions = regionprops(labeled_img)
|
|
num_objects = len(regions)
|
|
|
|
if num_objects == 0:
|
|
print(f"Tidak ditemukan objek pada gambar: {image_path}")
|
|
return None
|
|
|
|
# Inisialisasi variabel untuk agregasi
|
|
total_eccentricity = 0
|
|
total_metric = 0
|
|
total_area = 0
|
|
total_perimeter = 0
|
|
|
|
for region in regions:
|
|
# Fitur bentuk
|
|
eccentricity = calculate_eccentricity(region)
|
|
area = region.area
|
|
perimeter = region.perimeter
|
|
metric = (4 * np.pi * area) / (perimeter ** 2) if perimeter != 0 else 0
|
|
|
|
# Agregasi fitur
|
|
total_eccentricity += eccentricity
|
|
total_metric += metric
|
|
total_area += area
|
|
total_perimeter += perimeter
|
|
|
|
# Hitung rata-rata
|
|
mean_eccentricity = total_eccentricity / num_objects
|
|
mean_metric = total_metric / num_objects
|
|
|
|
# Membuat array fitur
|
|
features = np.array([[num_objects, mean_eccentricity, mean_metric, total_area, total_perimeter]])
|
|
|
|
return features
|
|
|
|
# Fungsi untuk melakukan prediksi
|
|
def predict(image_path, label_name):
|
|
features = extract_features(image_path, label_name)
|
|
if features is None:
|
|
return
|
|
|
|
# Melakukan prediksi
|
|
y_pred = rf_model.predict(features)
|
|
label_pred = le.inverse_transform(y_pred)
|
|
|
|
print(f"Prediksi untuk gambar {image_path}: {label_pred[0]}")
|
|
|
|
# Contoh penggunaan
|
|
image_path = 'D:/Skripsi/ekstraksi fitur/testing/Corynebacterium Diphteriae/cd100.jpg'
|
|
label_name = 'Corynebacterium Diphteriae' # Sesuaikan dengan label yang benar untuk preprocessing yang tepat
|
|
predict(image_path, label_name)
|