You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
item-sort/detect.py

377 lines
14 KiB
Python

# import the necessary packages
#from imutils import perspective
from imutils import contours
import numpy as np
import argparse
import imutils
import cv2
import math
import time
itemw = 0
itemh = 0
def midpoint(ptA, ptB):
return ((ptA[0] + ptB[0]) * 0.5, (ptA[1] + ptB[1]) * 0.5)
def sizeVexScrew(iteml):
# Screw Sizing code
# subtract screw head size to find thread length
shead = 0.1
iteml -= shead
#print("Thread Length: " + str(iteml))
iteml *= 8
iteml = round(iteml)
iteml /= 8
return iteml
def sizeStandoff(iteml):
# Standoff Sizing code
iteml *= 2
iteml = round(iteml)
iteml /= 2
return iteml
def larger(a, b):
if a >= b:
return a
else:
return b
def smaller(a, b):
if a < b:
return a
else:
return b
def near(a, b, close):
if abs(a-b) < close:
return True
return False
def swap(a, b):
tmp = a
a = b
b = tmp
"""
# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image", required=True,
help="path to the input image")
#ap.add_argument("-c", "--cascade", required=True,
# help="path to the cascade")
ap.add_argument("-w", "--width", type=float, required=True,
help="width of the left-most object in the image (in inches)")
ap.add_argument("-n", "--number", type=int, required=False,
help="object # to measure (from left to right)")
ap.add_argument("-s", "--show", action="store_true",
help="show on the screen")
args = vars(ap.parse_args())
args2 = ap.parse_args()"""
def detect(calibration_width, img_file, show, quick):
list = []
#if type(args["number"]) == type(selected):
# selected = args["number"]
# load the image, convert it to grayscale, and blur it slightly
image = None
#print(str(type(img_file)))
if str(type(img_file)) == "<class 'numpy.ndarray'>":
image = img_file.copy()
else:
image = cv2.imread(img_file)
#image = img_file.copy()
image = cv2.resize(image, (math.floor(image.shape[1]*0.5), math.floor(image.shape[0]*0.5)))
#image = cv2.resize(image, (1000, int(image.shape[0]/image.shape[1] * 1000)), interpolation=cv2.INTER_NEAREST)
if show and not quick:
cv2.namedWindow("Item Sorter")
cv2.imshow("Item Sorter", image)
cv2.waitKey(0)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (7, 7), 0)
gray = cv2.GaussianBlur(gray, (7, 7), 0)
# perform edge detection, then perform a dilation + erosion to
# close gaps in between object edges
edged = cv2.Canny(gray, 50, 100)
edged = cv2.dilate(edged, None, iterations=2)
edged = cv2.erode(edged, None, iterations=2)
edged = cv2.dilate(edged, None, iterations=1)
edged = cv2.erode(edged, None, iterations=1)
edged = cv2.dilate(edged, None, iterations=1)
#edged = cv2.erode(edged, None, iterations=1)
#edged = cv2.dilate(edged, None, iterations=1)
if show and not quick:
cv2.imshow("Item Sorter", edged)
cv2.waitKey(0)
# find contours in the edge map
cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
# sort the contours from left-to-right and initialize the
# 'pixels per metric' calibration variable
#(cnts, _) = contours.sort_contours(cnts)
pixelsPerMetric = None
num = 0
# Calibration loop
"""
for c in cnts:
# if the contour is not sufficiently large, ignore it
if cv2.contourArea(c) < 100:
continue
# compute the rotated bounding box of the contour
orig = image.copy()
box = cv2.minAreaRect(c)
# xpos,ypos,w,h = cv2.boundingRect(c)
# crop_img = orig[ypos:ypos+h, xpos:xpos+w]
# cv2.imwrite("object_images/IMG_" + str(w*h) + ".jpg", crop_img) # create training images
box = cv2.cv.BoxPoints(box) if imutils.is_cv2() else cv2.boxPoints(box)
box = np.array(box, dtype="int")
#box = perspective.order_points(box)
(tl, tr, br, bl) = box
(tltrX, tltrY) = midpoint(tl, tr)
(blbrX, blbrY) = midpoint(bl, br)
(tlblX, tlblY) = midpoint(tl, bl)
(trbrX, trbrY) = midpoint(tr, br)
dA = np.linalg.norm(np.array((tltrX, tltrY, 0)) -
np.array((blbrX, blbrY, 0)))
dB = np.linalg.norm(np.array((tlblX, tlblY, 0)) -
np.array((trbrX, trbrY, 0)))
area_box = dA * dB
(x, y), radius = cv2.minEnclosingCircle(c)
area_contour = cv2.contourArea(c)
area_circle = math.pi * pow(radius, 2)
boxiness = area_contour / area_box
circleness = area_contour / area_circle
circular = False
rectangular = False
if boxiness > circleness:
rectangular = True
cv2.drawContours(orig, [box.astype("int")], -1, (0, 255, 0), 2)
else:
circular = True
cv2.circle(orig, (int(x), int(y)), int(radius), (0, 255, 0), 3)
mask = np.zeros(gray.shape, np.uint8)
cv2.drawContours(mask, [c], 0, 255, -1)
#pixelpoints = np.transpose(np.nonzero(mask))
hsv = cv2.cvtColor(orig, cv2.COLOR_BGR2HSV)
mean_val = cv2.mean(hsv, mask=mask)
#print(str(mean_val[0]))
#print(", " + str(mean_val[0]/mean_val[2]))
#print(", " + str(mean_val[2]/mean_val[1]))
if pixelsPerMetric is None and circular is True and near(mean_val[0], 16, 4.5):
# and near(mean_val[0], 63, 40) is True and near(mean_val[1], 108, 40) is True and near(mean_val[2], 104, 40) is True:
pixelsPerMetric = smaller(dA, dB) / calibration_width
continue
"""
pixelsPerMetric = 25
orig = image.copy()
objtype = "Unknown"
objname = ""
# loop over the contours individually
for c in cnts:
#orig = image.copy()
num += 1
# if the contour is not sufficiently large, ignore it
#pixelsPerMetric = 75
if cv2.contourArea(c) < 300 or pixelsPerMetric is None:
continue
# compute the rotated bounding box of the contour
box = cv2.minAreaRect(c)
box = cv2.cv.BoxPoints(box) if imutils.is_cv2() else cv2.boxPoints(box)
box = np.array(box, dtype="int")
# unpack the ordered bounding box, then compute the midpoint
# between the top-left and top-right coordinates, followed by
# the midpoint between bottom-left and bottom-right coordinates
(tl, tr, br, bl) = box
(tltrX, tltrY) = midpoint(tl, tr)
(blbrX, blbrY) = midpoint(bl, br)
# compute the midpoint between the top-left and top-right points,
# followed by the midpoint between the top-right and bottom-right
(tlblX, tlblY) = midpoint(tl, bl)
(trbrX, trbrY) = midpoint(tr, br)
# compute the Euclidean distance between the midpoints
dA = np.linalg.norm(np.array((tltrX, tltrY, 0)) -
np.array((blbrX, blbrY, 0)))
dB = np.linalg.norm(np.array((tlblX, tlblY, 0)) -
np.array((trbrX, trbrY, 0)))
dimA = dA / pixelsPerMetric
dimB = dB / pixelsPerMetric
# Item detection
area_box = dA * dB
(x, y), radius = cv2.minEnclosingCircle(c)
area_contour = cv2.contourArea(c)
area_circle = math.pi * pow(radius, 2)
boxiness = area_contour / area_box
circleness = area_contour / area_circle
circular = False
rectangular = False
if boxiness > circleness:
rectangular = True
#cv2.drawContours(orig, [box.astype("int")], -1, (0, 255, 0), 2)
else:
circular = True
cv2.circle(orig, (int(x), int(y)), int(radius), (255, 0, 0), 2)
objtype = "Unknown"
itemw = larger(dimA, dimB)
itemwr = itemw
itemwr *= 8
itemwr = round(itemwr)
itemwr /= 8
itemh = smaller(dimA, dimB)
itemhr = itemh
itemhr *= 16
itemhr = round(itemhr)
itemhr /= 16
if circular and itemwr == 0.75:
objtype = "Penny"
iteml = 0
"""else:
if circular and near(radius * 2 / pixelsPerMetric, 0.4, 0.03):
# Keps nut or spacer
objtype = "Spacer"
mask = np.zeros(gray.shape, np.uint8)
cv2.drawContours(mask, [c], 0, 255, -1)
#pixelpoints = np.transpose(np.nonzero(mask))
hsv = cv2.cvtColor(orig, cv2.COLOR_BGR2HSV)
mean_val = cv2.mean(hsv, mask=mask)
mean_rgb = cv2.mean(orig, mask=mask)
if near(mean_rgb[2], 59, 3) and near(mean_val[1], 85, 5): #and near(mean_val[2], 78, 5):
objtype = "Keps Nut"
print(str(mean_rgb[2]) + objtype + str(mean_val[1]))
elif circular and near(radius / pixelsPerMetric, 0.23, 0.02):
objtype = "Washer"
#print(str(radius * 2 / pixelsPerMetric) + objtype)
epsilon = 3 # 0.02*cv2.arcLength(c,True)
# print(str(epsilon))
approx = cv2.approxPolyDP(c, epsilon, True)
hull = cv2.convexHull(approx, returnPoints=False)
hull2 = cv2.convexHull(c)
defects = cv2.convexityDefects(c, hull)
#print(str(defects.size) + " match")
cv2.drawContours(orig, (hull2), -1, (0, 0, 255), 3)
cv2.drawContours(orig, (approx), -1, (255, 0, 0), 3)
convexness = area_contour / cv2.contourArea(hull2)
#print(str(convexness) + " % fill")
# if not cv2.isContourConvex(approx):
# if cv2.matchShapes(hull, c, 1, 0.0) > 1:
if defects is not None and defects.size > 5 and (convexness < 0.9 or boxiness < 0.75) and rectangular:
objtype = "Screw"
iteml = larger(dimA, dimB)
#print("Screw Length (RAW): " + str(iteml))
iteml = sizeVexScrew(radius * 2 / pixelsPerMetric)
#print("Rounded Length: " + str(iteml))
else:
if itemhr == 0.3125 and rectangular:
objtype = "Standoff"
iteml = sizeStandoff(itemw)
if itemhr == 0.1875 and rectangular:
objtype = "Axle"
iteml = (radius * 2 / pixelsPerMetric + itemw) / 2
"""
rows, cols = orig.shape[:2]
[vx, vy, xx, yy] = cv2.fitLine(c, cv2.DIST_L2, 0, 0.01, 0.01)
lefty = int((-xx*vy/vx) + yy)
righty = int(((cols-xx)*vy/vx)+yy)
# cv2.line(orig,(cols-1,righty),(0,lefty),(0,255,0),2)
slope = (lefty - righty) / (1 - cols)
angle = math.atan(slope)
xpos = x - math.cos(angle) * radius
ypos = y - math.sin(angle) * radius
xpos2 = x + math.cos(angle) * radius
ypos2 = y + math.sin(angle) * radius
if xpos > xpos2:
swap(xpos, xpos2)
swap(ypos, ypos2)
if rectangular:
cv2.line(orig, (int(xpos), int(ypos)),
(int(xpos2), int(ypos2)), (255, 127, 0), 2)
# print(str(iteml))
# draw the object sizes on the image
# cv2.putText(orig, "{:.5f}in".format(itemhr),
# (int(trbrX + 20), int(trbrY)), cv2.FONT_HERSHEY_SIMPLEX,
# 0.65, (255, 255, 255), 2)
if objtype != "Penny":
objtype = magicSort(c)
if objtype == "Unknown":
objtype = magicSort(c)
output = "{:.2f}in".format(itemw) + " x {:.2f}in".format(itemh)
if circular:
cv2.putText(orig, str(objtype),
(int(x - 25), int(y + radius + 20)
), cv2.FONT_HERSHEY_SIMPLEX,
0.6, (50, 50, 220), 2)
else:
cv2.putText(orig, str(objtype),
(int(xpos2 + 10), int(ypos2 + 20)
), cv2.FONT_HERSHEY_SIMPLEX,
0.6, (50, 50, 220), 2)
output = ""
objname = objtype;
"""
if objtype == "Screw" or objtype == "Standoff":
output = str(iteml) + "in"
objname += str(iteml)
if objtype == "Axle":
output = "{:.2f}in".format(iteml)
objname += str(itemwr)
#print(objname)
"""
list.append(objname)
if circular:
cv2.putText(orig, output, # print data
(int(x - 25), int(y + radius + 40)
), cv2.FONT_HERSHEY_SIMPLEX,
0.5, (50, 50, 220), 1)
else:
cv2.putText(orig, output, # print data
(int(xpos2 + 10), int(ypos2 + 40)
), cv2.FONT_HERSHEY_SIMPLEX,
0.5, (50, 50, 220), 1)
# show the output image
if show and not quick:
cv2.imshow("Item Sorter", orig)
#cv2.waitKey(1)
if not quick:
cv2.waitKey(0)
return (list, orig)
def magicSort(contour):
moments = cv2.moments(contour)
humoments = cv2.HuMoments(moments)
#humoments[6] = abs(humoments[6]) #it's possible for the last number to change sign if item is mirrored
#magicNumber1 = 0
#magicNumber2 = 0
name = "Unknown"
for i in range(0,7):
if humoments[i] == 0:
humoments[i] = 0.1;
humoments[i] = -1 * math.copysign(1.0, humoments[i]) * math.log10(abs(humoments[i]))
if i > 1:
humoments[i] = int(round(humoments[i][0] / 8) * 8)
if i != 4 and i != 6:
name += ", " + str(abs(int(humoments[i][0])))
#magicNumber1 += abs(humoments[i][0])
else:
humoments[i] = int(round(humoments[i][0] * 4) * 16)
name += ", " + str(abs(int(humoments[i][0])))
#magicNumber2 += abs(humoments[i][0])
#magicNumber += humoments[i][0]
#print(str(humoments))
#print(magicNumber)
#name = "Unknown: " + str(int(magicNumber1)) + ", " + str(int(magicNumber2))
#print(name)
return name