[https://drive.google.com/file/d/1fUH892Q6hbfDnoXF1G3-ZPe_rFPCamPn/view?usp=sharing]
[https://drive.google.com/file/d/17ZiHIFTwMApmttT1HzDMasVGDgNlUJhM/view?usp=sharing]Actually我正在使用开放式cv python进行异物检测。我正在捕获第一帧并减去所有其余帧，因此，只要第一帧发生变化，它将检测为异物。从第一帧中减去帧后，我发现轮廓并找出轮廓，只要在特定数量的帧（由连续帧设置为5）中，特定轮廓保留在帧中，它将检测为异物。因此，帧中缺少的某些东西也改变了前景，这就是为什么它也检测到了这一点。我需要一种解决方案，因为它只能检测添加到框架中的东西。请帮帮我！
实际上，我正在使用开放cv python进行异物检测。我正在捕获第一帧并减去所有其余帧，因此，只要第一帧发生变化，它将检测为异物。
import numpy as np
import cv2
from collections import Counter, defaultdict
import imutils
#file_path ='0'
cap = cv2.VideoCapture(0)
grabbed, frame = cap.read()
#frame = imutils.resize(frame,width =500)
img = cv2.imwrite("shot.jpg",frame)
firstframe_path ='shot.jpg'
firstframe = cv2.imread(firstframe_path)
firstframe_blur = cv2.GaussianBlur(firstframe,(21,21),0)
firstframe_gray = cv2.cvtColor(firstframe_blur, cv2.COLOR_BGR2GRAY)
cv2.namedWindow('CannyEdgeDet',cv2.WINDOW_NORMAL)
cv2.namedWindow('Abandoned Object Detection',cv2.WINDOW_NORMAL)
cv2.namedWindow('Morph_CLOSE',cv2.WINDOW_NORMAL)
cap = cv2.VideoCapture(0)
consecutiveframe=5
track_temp=[]
track_master=[]
track_temp2=[]
top_contour_dict = defaultdict(int)
obj_detected_dict = defaultdict(int)
frameno = 0
# roi = cv2.selectROI(frame,showCrosshair=False)
# cv2.destroyWindow('ROI selector')
# firstframe = firstframe[roi[1]:roi[1]+roi[3],roi[0]:roi[0]+roi[2]]
while (cap.isOpened()):
ret, frame = cap.read()
#frame = imutils.resize(frame,width =500)
#roi = cv2.selectROI(frame,showCrosshair=False)
#forig = frame.copy()
#frame=frame[roi[1]:roi[1]+roi[3],roi[0]:roi[0]+roi[2]]
if ret==0:
break
frameno = frameno + 1
#cv2.putText(frame,'%s%.f'%('Frameno:',frameno),(50,50),cv2.FONT_HERSHEY_SIMPLEX, 1, (255,255,255),2)
frame_diff = cv2.absdiff(firstframe, frame)
cv2.imshow('frame_diff',frame_diff)
frame_gray = cv2.cvtColor(frame_diff, cv2.COLOR_BGR2GRAY)
frame_blur = cv2.GaussianBlur(frame_gray,(21,21),0)
edged = cv2.Canny(frame_blur,10,200)
cv2.imshow('CannyEdgeDet',edged)
kernel2 = np.ones((5,5),np.uint8)
thresh2 = cv2.morphologyEx(edged,cv2.MORPH_CLOSE,
kernel2,iterations=2)
#cv2.imshow('Morph_Close', thresh2)
cnts,_ = cv2.findContours(thresh2.copy(),cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
mycnts =[]
for c in cnts:
M = cv2.moments(c)
if M['m00'] == 0:
pass
else:
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
if cv2.contourArea(c) < 300: ##or cv2.contourArea(c)>20000:
pass
else:
mycnts.append(c)
(x, y, w, h) = cv2.boundingRect(c)
sumcxcy=cx+cy
#print(sumcxcy)
#track_list.append(cx+cy)
track_temp.append([cx+cy,frameno])
track_master.append([cx+cy,frameno])
#print("did:",did,frameno)
countuniqueframe = set(j for i, j in track_master) # get a #set of unique frameno. then len(countuniqueframe)
#print("pip:",countuniqueframe)
#-------------------------------------------------------
---------
# Store history of frames ; no. of frames stored set by
#'consecutiveframe' ;
# if no. of no. of unique frames > consecutiveframes,
#then 'pop or remove' the earliest frame ; defined by
# minframeno. Objective is to count the same values
#occurs in all the frames under this list. if yes,
# it is likely that it is a stationary object and not a
#passing object (walking)
# And the value is stored separately in
#top_contour_dict , and counted each time. This dict is
#the master
# dict to store the list of suspecious object. Ideally,
#it should be a short list. if there is a long list
# there will be many false detection. To keep the list
#short, increase the 'consecutiveframe'.
# Keep the number of frames to , remove the
#minframeno.; but hard to remove, rather form a new list
#without
#the minframeno.
#------------------------------------------------------
----------
if len(countuniqueframe)>consecutiveframe or False:
minframeno=min(j for i, j in track_master)
for i, j in track_master:
if j != minframeno: # get a new list. omit the those with the minframeno
track_temp2.append([i,j])
track_master=list(track_temp2) # transfer to the
master list
track_temp2=[]
#print 'After',track_master
#count each of the sumcxcy
#if the same sumcxcy occurs in all the frames, store in master contour dictionary, add 1
countcxcy = Counter(i for i, j in track_master)
#print countcxcy
#example countcxcy : Counter({544: 1, 537: 1, 530: 1,
#523: 1, 516: 1})
#if j which is the count occurs in all the frame, store
#the sumcxcy in dictionary, add 1
for i,j in countcxcy.items():
if j>=consecutiveframe:
top_contour_dict[i] += 1
if sumcxcy in top_contour_dict:
if top_contour_dict[sumcxcy]>50:
cv2.rectangle(frame, (x, y), (x + w, y + h),(255, 0, 0), 3)
cv2.putText(frame,'%s'%('foreign_Object'),(cx,cy),cv2.FONT_HERSHEY_SIMPLEX, 0.6,(255,255,255),2)
#cv2.putText(frame, "foreign object !!!", (10,20),cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2)
print ('Detected : ', sumcxcy,frameno,obj_detected_dict)
# Store those objects that are detected, and store
#the last frame that it happened.
# Need to find a way to clean the top_contour_dict,
#else contour will be detected after the
# object is removed because the value is still in
#the dict.
# Method is to record the last frame that the object
#is detected with the Current Frame (frameno)
# if Current Frame - Last Frame detected > some big
#number say 100 x 3, then it means that
# object may have been removed because it has not
#been detected for 100x3 frames.
obj_detected_dict[sumcxcy]=frameno
for i, j in obj_detected_dict.items():
if frameno - obj_detected_dict[i]>200:
print ('PopBefore',i,obj_detected_dict[i],frameno,obj_detected_dict)
print ('PopBefore : top_contour:',top_contour_dict)obj_detected_dict.pop(i)
# Set the count for eg 448 to zero. because it has not be
#'activated' for 200 frames. Likely, to have been removed.
top_contour_dict[i]=0
print('PopAfter',i,obj_detected_dict[i],frameno,obj_detected_dict)
print ('PopAfter : top_contour :',top_contour_dict)
#cv2.rectangle(forig, (roi[0], roi[1]),(roi[0]+roi[2],roi[1]+roi[3]), (0, 255, 0),1)
cv2.imshow('Detections',frame)
if cv2.waitKey(25) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
我需要一种解决方案，因为它只能检测添加到框架中的东西。