Skip to content

Instantly share code, notes, and snippets.

@creotiv
Last active August 29, 2018 13:30
Show Gist options
  • Star 3 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save creotiv/febf149ae2211f70fd45c93b4b0218b1 to your computer and use it in GitHub Desktop.
Save creotiv/febf149ae2211f70fd45c93b4b0218b1 to your computer and use it in GitHub Desktop.
Class that used in car counting during CV traffic analysis
import logging
import math
import numpy as np
from scipy import spatial
import cv2
# ============================================================================
CAR_COLOURS = [ (0,0,255), (0,106,255), (0,216,255), (0,255,182), (0,255,76)
, (144,255,0), (255,255,0), (255,148,0), (255,0,178), (220,0,255) ]
class VehicleCounter(object):
def __init__(self, shape, divider):
self.log = logging.getLogger("vehicle_counter")
self.height, self.width = shape
self.divider = divider
self.vehicle_count = 0
self.path_size = 10
self.pathes = []
self.max_dst = 30
def distance(self, x,y,type='euclidian'):
if type == 'euclidian':
return math.sqrt((x[0]-y[0])**2+(x[1]-y[1])**2)
def update_count(self, matches, output_image = None):
self.log.debug("Updating count using %d matches...", len(matches))
self.update_pathes(matches)
self.draw_pathes(output_image)
cv2.putText(output_image, ("%02d" % self.vehicle_count), (150, 10)
, cv2.FONT_HERSHEY_PLAIN, 0.7, (127, 255, 255), 1)
self.log.debug("VEHICLES FOUND: %s" % self.vehicle_count)
def update_pathes(self, matches):
matches = np.array(matches)[:,-1]
matches = matches.tolist()
if not matches:
return
# add new points if pathes is empty
if not self.pathes:
for match in matches:
self.pathes.append([match])
else:
# link new points with old pathes based on minimum distance between points
count = len(matches)
new_pathes = []
for i,path in enumerate(self.pathes):
_min = 999999
_match = None
for match in matches:
d = self.distance(match,path[-1])
if d < _min:
_min = d
_match = match
if _match and _min <= self.max_dst:
matches.remove(_match)
path.append(_match)
# drop pathes that not linked
if _min <= self.max_dst:
new_pathes.append(path)
self.pathes = new_pathes
# add new pathes
if len(matches):
for match in matches:
# do not add matches that already should be counted
if match[1] > self.divider:
continue
self.pathes.append([match])
# save only last N points in path
for i,_ in enumerate(self.pathes):
self.pathes[i] = self.pathes[i][self.path_size*-1:]
# count vechiles and drop counted pathes:
new_pathes = []
for i,path in enumerate(self.pathes):
d = path[-2:]
if len(d) < 2:
new_pathes.append(path)
continue
if d[0][1] <= self.divider and d[1][1] > self.divider:
self.vehicle_count += 1
else:
new_pathes.append(path)
self.pathes = new_pathes
def draw_pathes(self, output_image=None):
if not output_image.any():
return
for i, path in enumerate(self.pathes):
for point in path:
cv2.circle(output_image, point, 2, CAR_COLOURS[i], -1)
cv2.polylines(output_image, [np.int32(path)]
, False, CAR_COLOURS[i], 1)
# ============================================================================
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment