今天是第五天，我們可以寫一個對於多隻斑馬魚行為分析的yolo程式了，以下是程式碼
import cv2
import numpy as np
from scipy.spatial import distance
# 載入 YOLO 模型
def load_yolo_model():
net = cv2.dnn.readNet("yolov3.weights", "yolov3.cfg")
with open("coco.names", "r") as f:
classes = [line.strip() for line in f.readlines()]
layer_names = net.getLayerNames()
output_layers = [layer_names[i - 1] for i in net.getUnconnectedOutLayers()]
return net, classes, output_layers
# 偵測斑馬魚
def detect_fish(frame, net, output_layers):
height, width, channels = frame.shape
blob = cv2.dnn.blobFromImage(frame, 0.00392, (416, 416), (0, 0, 0), True, crop=False)
net.setInput(blob)
outs = net.forward(output_layers)
class_ids = []
confidences = []
boxes = []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > 0.5 and class_id == 0: # 假設斑馬魚的類別ID是0
center_x = int(detection[0] * width)
center_y = int(detection[1] * height)
w = int(detection[2] * width)
h = int(detection[3] * height)
x = int(center_x - w / 2)
y = int(center_y - h / 2)
boxes.append([x, y, w, h])
confidences.append(float(confidence))
class_ids.append(class_id)
indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)
return boxes, confidences, indexes
# 繪製偵測結果
def draw_labels(boxes, confidences, indexes, frame):
font = cv2.FONT_HERSHEY_PLAIN
for i in range(len(boxes)):
if i in indexes:
x, y, w, h = boxes[i]
label = str(confidences[i])
color = (0, 255, 0)
cv2.rectangle(frame, (x, y), (x + w, y + h), color, 2)
cv2.putText(frame, label, (x, y - 10), font, 1, color, 1)
return frame
# 追蹤斑馬魚
def track_fish(boxes, indexes, frame, previous_positions):
current_positions = []
for i in range(len(boxes)):
if i in indexes:
x, y, w, h = boxes[i]
center_x = x + w // 2
center_y = y + h // 2
current_positions.append((center_x, center_y))
if previous_positions:
for prev in previous_positions:
distances = [distance.euclidean(prev, curr) for curr in current_positions]
min_dist_index = np.argmin(distances)
min_dist = distances[min_dist_index]
if min_dist < 50: # 設定一個閾值來判斷是否為同一隻斑馬魚
cv2.line(frame, prev, current_positions[min_dist_index], (255, 0, 0), 2)
return current_positions
# 主程式
def main():
net, classes, output_layers = load_yolo_model()
cap = cv2.VideoCapture("zebrafish_video.mp4")
previous_positions = []
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
boxes, confidences, indexes = detect_fish(frame, net, output_layers)
frame = draw_labels(boxes, confidences, indexes, frame)
current_positions = track_fish(boxes, indexes, frame, previous_positions)
previous_positions = current_positions
cv2.imshow("Zebrafish Detection and Tracking", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
if __name__ == "__main__":
main()
def load_yolo_model():
net = cv2.dnn.readNet("yolov3.weights", "yolov3.cfg")
with open("coco.names", "r") as f:
classes = [line.strip() for line in f.readlines()]
layer_names = net.getLayerNames()
output_layers = [layer_names[i - 1] for i in net.getUnconnectedOutLayers()]
return net, classes, output_layers
這個函數的工作是把 YOLO 模型載入進來。我們需要提供 YOLO 的配置文件和權重文件，還有一個包含物體名稱的文件。這樣我們才能使用 YOLO 來偵測物體。我們還要知道哪些是 YOLO 的輸出層，這樣我們才能得到偵測結果。
def detect_fish(frame, net, output_layers):
height, width, channels = frame.shape
blob = cv2.dnn.blobFromImage(frame, 0.00392, (416, 416), (0, 0, 0), True, crop=False)
net.setInput(blob)
outs = net.forward(output_layers)
class_ids = []
confidences = []
boxes = []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > 0.5 and class_id == 0: # 假設斑馬魚的類別ID是0
center_x = int(detection[0] * width)
center_y = int(detection[1] * height)
w = int(detection[2] * width)
h = int(detection[3] * height)
x = int(center_x - w / 2)
y = int(center_y - h / 2)
boxes.append([x, y, w, h])
confidences.append(float(confidence))
class_ids.append(class_id)
indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)
return boxes, confidences, indexes
這個函數用來偵測斑馬魚。它會先把影像轉換成 YOLO 可以處理的格式，然後把它丟進 YOLO 模型裡進行偵測。偵測出來的結果會包含每個物體的位置和信心度（即 YOLO 認為它是斑馬魚的可能性）。我們只保留信心度超過 50% 的結果。
def draw_labels(boxes, confidences, indexes, frame):
font = cv2.FONT_HERSHEY_PLAIN
for i in range(len(boxes)):
if i in indexes:
x, y, w, h = boxes[i]
label = str(confidences[i])
color = (0, 255, 0)
cv2.rectangle(frame, (x, y), (x + w, y + h), color, 2)
cv2.putText(frame, label, (x, y - 10), font, 1, color, 1)
return frame
這個函數用來在影像上畫出偵測結果。它會在每個偵測到的斑馬魚周圍畫一個框，並在框上方顯示信心度。這樣我們就能看到 YOLO 偵測到的斑馬魚。
def track_fish(boxes, indexes, frame, previous_positions):
current_positions = []
for i in range(len(boxes)):
if i in indexes:
x, y, w, h = boxes[i]
center_x = x + w // 2
center_y = y + h // 2
current_positions.append((center_x, center_y))
if previous_positions:
for prev in previous_positions:
distances = [distance.euclidean(prev, curr) for curr in current_positions]
min_dist_index = np.argmin(distances)
min_dist = distances[min_dist_index]
if min_dist < 50: # 設定一個閾值來判斷是否為同一隻斑馬魚
cv2.line(frame, prev, current_positions[min_dist_index], (255, 0, 0), 2)
return current_positions
這個函數用來追蹤斑馬魚。它會記錄每幀影像中斑馬魚的位置，然後比較當前幀和上一幀的位置，看看哪些斑馬魚可能是同一隻。如果兩個位置之間的距離小於 50 像素，我們就認為它們是同一隻斑馬魚，並畫出它們的運動軌跡。
def main():
net, classes, output_layers = load_yolo_model()
cap = cv2.VideoCapture("zebrafish_video.mp4")
previous_positions = []
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
boxes, confidences, indexes = detect_fish(frame, net, output_layers)
frame = draw_labels(boxes, confidences, indexes, frame)
current_positions = track_fish(boxes, indexes, frame, previous_positions)
previous_positions = current_positions
cv2.imshow("Zebrafish Detection and Tracking", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
if __name__ == "__main__":
main()
這是主程式。它首先載入 YOLO 模型，然後打開一個包含斑馬魚的影片檔案。每讀取一幀影像，我們就用 YOLO 模型進行偵測，畫出偵測結果，再追蹤斑馬魚的運動軌跡。最後顯示處理後的影像。如果按下 'q' 鍵，程式就會結束。
總結來說，這段程式碼是用來實時偵測並追蹤影片中的斑馬魚，讓我們可以觀察它們的行為。
iThome鐵人賽
看影片追技術