package/heatmap
1
0
Fork 0
mirror of https://github.com/kerberos-io/heatmap.git synced 2025-09-26 20:31:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

cleaned up the file

Browse Source
This commit is contained in:
QuintenMM
2022-01-25 14:21:12 +01:00
parent 0499d98db1
commit ed247309f9
1 changed files with 133 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

235
person_tracker.py
View File

@@ -1,9 +1,11 @@
import cv2 import cv2
import datetime import datetime
import imutils
import numpy as np import numpy as np
from collections import defaultdict from collections import defaultdict
from centroidtracker import CentroidTracker from centroidtracker import CentroidTracker
import pandas as pd
import imutils
protopath = "MobileNetSSD_deploy.prototxt" protopath = "MobileNetSSD_deploy.prototxt"
modelpath = "MobileNetSSD_deploy.caffemodel" modelpath = "MobileNetSSD_deploy.caffemodel"
@@ -15,141 +17,170 @@ detector = cv2.dnn.readNetFromCaffe(prototxt=protopath, caffeModel=modelpath)
CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat", CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat",
"bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable",
"dog", "horse", "motorbike", "person", "pottedplant", "sheep", "dog", "horse", "motorbike", "person", "pottedplant", "sheep",
"sofa", "train", "tvmonitor"] "sofa", "train", "tvmonitor"]
# maxDisappeared, time wait when object moves out of frame # maxDisappeared, time wait when object moves out of frame
tracker = CentroidTracker(maxDisappeared=700, maxDistance=220) tracker = CentroidTracker(maxDisappeared=500, maxDistance=220)
def non_max_suppression_fast(boxes, overlapThresh): def non_max_suppression_fast(boxes, overlapThresh):
try: '"Cobine boundingboxes that overlap into one bbox"'
if len(boxes) == 0: try:
return [] if len(boxes) == 0:
return []
if boxes.dtype.kind == "i": if boxes.dtype.kind == "i":
boxes = boxes.astype("float") boxes = boxes.astype("float")
pick = [] pick = []
x1 = boxes[:, 0] x1 = boxes[:, 0]
y1 = boxes[:, 1] y1 = boxes[:, 1]
x2 = boxes[:, 2] x2 = boxes[:, 2]
y2 = boxes[:, 3] y2 = boxes[:, 3]
area = (x2 - x1 + 1) * (y2 - y1 + 1)
idxs = np.argsort(y2)
area = (x2 - x1 + 1) * (y2 - y1 + 1) while len(idxs) > 0:
idxs = np.argsort(y2) last = len(idxs) - 1
i = idxs[last]
pick.append(i)
while len(idxs) > 0: xx1 = np.maximum(x1[i], x1[idxs[:last]])
last = len(idxs) - 1 yy1 = np.maximum(y1[i], y1[idxs[:last]])
i = idxs[last] xx2 = np.minimum(x2[i], x2[idxs[:last]])
pick.append(i) yy2 = np.minimum(y2[i], y2[idxs[:last]])
xx1 = np.maximum(x1[i], x1[idxs[:last]]) w = np.maximum(0, xx2 - xx1 + 1)
yy1 = np.maximum(y1[i], y1[idxs[:last]]) h = np.maximum(0, yy2 - yy1 + 1)
xx2 = np.minimum(x2[i], x2[idxs[:last]])
yy2 = np.minimum(y2[i], y2[idxs[:last]])
w = np.maximum(0, xx2 - xx1 + 1) overlap = (w * h) / area[idxs[:last]]
h = np.maximum(0, yy2 - yy1 + 1)
overlap = (w * h) / area[idxs[:last]] idxs = np.delete(idxs, np.concatenate(([last],
np.where(overlap > overlapThresh)[0])))
idxs = np.delete(idxs, np.concatenate(([last], return boxes[pick].astype("int")
np.where(overlap > overlapThresh)[0]))) except Exception as e:
print("Exception occurred in non_max_suppression : {}".format(e))
return boxes[pick].astype("int") def convert_to_2d(Xcenter, y2):
except Exception as e: '" Convert the coordinates from a 3d playing field to a 2d playing field"'
print("Exception occurred in non_max_suppression : {}".format(e)) pts_src = np.array([[257, 262], [370, 225], [492, 190], [294, 324], [474, 272], [620, 213], [727, 383], [799, 259]])
# Take points from the frame as reference and give the same point coordinates on the picture for a transformation
pts_dst = np.array([[110, 145], [349, 145], [588, 145], [110, 500], [349, 500], [588, 500], [349, 855], [588, 855]])
# calculate matrix H
h, status = cv2.findHomography(pts_src, pts_dst)
def main(): # provide a point you wish to map from image 1 to image 2
cap = cv2.VideoCapture('1639943552_6-967003_camera1_200-200-400-400_24_769.mp4') a = np.array([[Xcenter, y2]], dtype='float32')
a = np.array([a])
fps_start_time = datetime.datetime.now() # finally, get the mapping
fps = 0 pointsOut = cv2.perspectiveTransform(a, h)
total_frames = 0 pointsOut = pointsOut.astype(int)
centroid_dict = defaultdict(list) return pointsOut
object_id_list = []
while True: def main(video="1639943552_6-967003_camera1_200-200-400-400_24_769.mp4"):
ret, frame = cap.read() '"Read the frames, recognise humans and track them. The coordinates of the bottom of the bbox are saved for transormation and plotting"'
frame = imutils.resize(frame, width=600) cap = cv2.VideoCapture(video)
total_frames = total_frames + 1
(height, width) = frame.shape[:2]
blob = cv2.dnn.blobFromImage(frame, 0.007843, (width, height), 127.5) paddel_2d = cv2.imread('media/paddelfield.jpeg')
detector.setInput(blob)
person_detections = detector.forward()
rects = [] fps_start_time = datetime.datetime.now()
for i in np.arange(0, person_detections.shape[2]): fps = 0
confidence = person_detections[0, 0, i, 2] total_frames = 0
if confidence > 0.5: centroid_dict = defaultdict(list)
idx = int(person_detections[0, 0, i, 1]) object_id_list = []
if CLASSES[idx] != "person": while True:
continue ret, frame = cap.read()
frame = imutils.resize(frame, width=800)
# frame = frame[540:1080, 700:1920]
total_frames = total_frames + 1
(height, width) = frame.shape[:2]
person_box = person_detections[0, 0, i, 3:7] * np.array([width, height, width, height]) blob = cv2.dnn.blobFromImage(frame, 0.007843, (width, height), 127.5)
(startX, startY, endX, endY) = person_box.astype("int") detector.setInput(blob)
rects.append(person_box) person_detections = detector.forward()
boundingboxes = np.array(rects) rects = []
boundingboxes = boundingboxes.astype(int) for i in np.arange(0, person_detections.shape[2]):
rects = non_max_suppression_fast(boundingboxes, 0.4) confidence = person_detections[0, 0, i, 2]
objects = tracker.update(rects) if confidence > 0.5:
idx = int(person_detections[0, 0, i, 1])
for (objectId, bbox) in objects.items(): if CLASSES[idx] != "person":
x1, y1, x2, y2 = bbox continue
x1 = int(x1)
y1 = int(y1)
x2 = int(x2)
y2 = int(y2)
xCenter = int((x1 + x2) / 2) person_box = person_detections[0, 0, i, 3:7] * np.array([width, height, width, height])
yCenter = int((y1 + y2) / 2) (startX, startY, endX, endY) = person_box.astype("int")
rects.append(person_box)
cv2.circle(frame, (xCenter, y2), 5, (0, 255, 0), -1) boundingboxes = np.array(rects)
boundingboxes = boundingboxes.astype(int)
rects = non_max_suppression_fast(boundingboxes, 0.4)
objects = tracker.update(rects)
centroid_dict[objectId].append((xCenter, y2)) for (objectId, bbox) in objects.items():
x1, y1, x2, y2 = bbox
x1 = int(x1)
y1 = int(y1)
x2 = int(x2)
y2 = int(y2)
if objectId not in object_id_list: xCenter = int((x1 + x2) / 2)
object_id_list.append(objectId) yCenter = int((y1 + y2) / 2)
start_pt = (xCenter, y2)
end_pt = (xCenter, y2)
cv2.line(frame, start_pt, end_pt, (0, 255, 0), 2)
else:
L = len(centroid_dict[objectId])
for pt in range(len(centroid_dict[objectId])):
if not pt + 1 == L:
start_pt = (centroid_dict[objectId][pt][0], centroid_dict[objectId][pt][1])
end_pt = (centroid_dict[objectId][pt + 1][0], centroid_dict[objectId][pt + 1][1])
cv2.line(frame, start_pt, end_pt, (0, 255, 0), 1)
cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 2) cv2.circle(frame, (xCenter, y2), 5, (0, 255, 0), -1)
text = "ID: {}".format(objectId)
cv2.putText(frame, text, (x1, y1-5), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 1)
fps_end_time = datetime.datetime.now() # saving the converted cords
time_diff = fps_end_time - fps_start_time pointsout = convert_to_2d(xCenter, y2)
if time_diff.seconds == 0: for tuple in pointsout:
fps = 0.0 for points in tuple:
else: pd.DataFrame({'x': [points[0]], 'y': [points[1]]}, index=[objectId]).to_csv('cords.csv', mode='a',
fps = (total_frames / time_diff.seconds) header=False)
fps_text = "FPS: {:.2f}".format(fps)
cv2.putText(frame, fps_text, (5, 30), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 1)
cv2.imshow("Application", frame) centroid_dict[objectId].append((xCenter, y2))
# cv2.VideoWriter_fourcc("new_vid.mp4", cv2.VideoWriter_fourcc(*'mp4v'), 20, (width, height))
key = cv2.waitKey(1) if objectId not in object_id_list:
if key == ord('q'): object_id_list.append(objectId)
break start_pt = (xCenter, y2)
cap.release() end_pt = (xCenter, y2)
cv2.destroyAllWindows() cv2.line(frame, start_pt, end_pt, (0, 255, 0), 2)
else:
L = len(centroid_dict[objectId])
for pt in range(len(centroid_dict[objectId])):
if not pt + 1 == L:
start_pt = (centroid_dict[objectId][pt][0], centroid_dict[objectId][pt][1])
end_pt = (centroid_dict[objectId][pt + 1][0], centroid_dict[objectId][pt + 1][1])
cv2.line(frame, start_pt, end_pt, (0, 255, 0), 1)
cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 2)
text = "ID: {}".format(objectId)
cv2.putText(frame, text, (x1, y1 - 5), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 1)
fps_end_time = datetime.datetime.now()
time_diff = fps_end_time - fps_start_time
if time_diff.seconds == 0:
fps = 0.0
else:
fps = (total_frames / time_diff.seconds)
fps_text = "FPS: {:.2f}".format(fps)
cv2.putText(frame, fps_text, (5, 30), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 1)
# cv2.imshow("Application", frame)
key = cv2.waitKey(1)
if key == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
main() main()