predict.py

import cv2
import numpy as np
from numpy.linalg import norm
import sys
import os
import json

SZ = 20          #训练图片长宽
MAX_WIDTH = 1000 #原始图片最大宽度
Min_Area = 2000  #车牌区域允许最大面积
PROVINCE_START = 1000
#读取图片文件
def imreadex(filename):
	return cv2.imdecode(np.fromfile(filename, dtype=np.uint8), cv2.IMREAD_COLOR)
	
def point_limit(point):
	if point[0] < 0:
		point[0] = 0
	if point[1] < 0:
		point[1] = 0

#根据设定的阈值和图片直方图，找出波峰，用于分隔字符
def find_waves(threshold, histogram):
	up_point = -1#上升点
	is_peak = False
	if histogram[0] > threshold:
		up_point = 0
		is_peak = True
	wave_peaks = []
	for i,x in enumerate(histogram):
		if is_peak and x < threshold:
			if i - up_point > 2:
				is_peak = False
				wave_peaks.append((up_point, i))
		elif not is_peak and x >= threshold:
			is_peak = True
			up_point = i
	if is_peak and up_point != -1 and i - up_point > 4:
		wave_peaks.append((up_point, i))
	return wave_peaks

#根据找出的波峰，分隔图片，从而得到逐个字符图片
def seperate_card(img, waves):
	part_cards = []
	for wave in waves:
		part_cards.append(img[:, wave[0]:wave[1]])
	return part_cards

#来自opencv的sample，用于svm训练
def deskew(img):
	m = cv2.moments(img)
	if abs(m['mu02']) < 1e-2:
		return img.copy()
	skew = m['mu11']/m['mu02']
	M = np.float32([[1, skew, -0.5*SZ*skew], [0, 1, 0]])
	img = cv2.warpAffine(img, M, (SZ, SZ), flags=cv2.WARP_INVERSE_MAP | cv2.INTER_LINEAR)
	return img
#来自opencv的sample，用于svm训练
def preprocess_hog(digits):
	samples = []
	for img in digits:
		gx = cv2.Sobel(img, cv2.CV_32F, 1, 0)
		gy = cv2.Sobel(img, cv2.CV_32F, 0, 1)
		mag, ang = cv2.cartToPolar(gx, gy)
		bin_n = 16
		bin = np.int32(bin_n*ang/(2*np.pi))
		bin_cells = bin[:10,:10], bin[10:,:10], bin[:10,10:], bin[10:,10:]
		mag_cells = mag[:10,:10], mag[10:,:10], mag[:10,10:], mag[10:,10:]
		hists = [np.bincount(b.ravel(), m.ravel(), bin_n) for b, m in zip(bin_cells, mag_cells)]
		hist = np.hstack(hists)
		
		# transform to Hellinger kernel
		eps = 1e-7
		hist /= hist.sum() + eps
		hist = np.sqrt(hist)
		hist /= norm(hist) + eps
		
		samples.append(hist)
	return np.float32(samples)
#不能保证包括所有省份
provinces = [
"zh_cuan", "川",
"zh_e", "鄂",
"zh_gan", "赣",
"zh_gan1", "甘",
"zh_gui", "贵",
"zh_gui1", "桂",
"zh_hei", "黑",
"zh_hu", "沪",
"zh_ji", "冀",
"zh_jin", "津",
"zh_jing", "京",
"zh_jl", "吉",
"zh_liao", "辽",
"zh_lu", "鲁",
"zh_meng", "蒙",
"zh_min", "闽",
"zh_ning", "宁",
"zh_qing", "靑",
"zh_qiong", "琼",
"zh_shan", "陕",
"zh_su", "苏",
"zh_sx", "晋",
"zh_wan", "皖",
"zh_xiang", "湘",
"zh_xin", "新",
"zh_yu", "豫",
"zh_yu1", "渝",
"zh_yue", "粤",
"zh_yun", "云",
"zh_zang", "藏",
"zh_zhe", "浙"
]
class StatModel(object):
	def load(self, fn):
		self.model = self.model.load(fn)  
	def save(self, fn):
		self.model.save(fn)
class SVM(StatModel):
	def __init__(self, C = 1, gamma = 0.5):
		self.model = cv2.ml.SVM_create()
		self.model.setGamma(gamma)
		self.model.setC(C)
		self.model.setKernel(cv2.ml.SVM_RBF)
		self.model.setType(cv2.ml.SVM_C_SVC)
#训练svm
	def train(self, samples, responses):
		self.model.train(samples, cv2.ml.ROW_SAMPLE, responses)
#字符识别
	def predict(self, samples):
		r = self.model.predict(samples)
		return r[1].ravel()

class CardPredictor:
	def __init__(self):
		#车牌识别的部分参数保存在js中，便于根据图片分辨率做调整
		f = open('config.js')
		j = json.load(f)
		for c in j["config"]:
			if c["open"]:
				self.cfg = c.copy()
				break
		else:
			raise RuntimeError('没有设置有效配置参数')

	def __del__(self):
		self.save_traindata()
	def train_svm(self):
		#识别英文字母和数字
		self.model = SVM(C=1, gamma=0.5)
		#识别中文
		self.modelchinese = SVM(C=1, gamma=0.5)
		if os.path.exists("svm.dat"):
			self.model.load("svm.dat")
		else:
			chars_train = []
			chars_label = []
			
			for root, dirs, files in os.walk("train\\chars2"):
				if len(os.path.basename(root)) > 1:
					continue
				root_int = ord(os.path.basename(root))
				for filename in files:
					filepath = os.path.join(root,filename)
					digit_img = cv2.imread(filepath)
					digit_img = cv2.cvtColor(digit_img, cv2.COLOR_BGR2GRAY)
					chars_train.append(digit_img)
					#chars_label.append(1)
					chars_label.append(root_int)
			
			chars_train = list(map(deskew, chars_train))
			chars_train = preprocess_hog(chars_train)
			#chars_train = chars_train.reshape(-1, 20, 20).astype(np.float32)
			chars_label = np.array(chars_label)
			print(chars_train.shape)
			self.model.train(chars_train, chars_label)
		if os.path.exists("svmchinese.dat"):
			self.modelchinese.load("svmchinese.dat")
		else:
			chars_train = []
			chars_label = []
			for root, dirs, files in os.walk("train\\charsChinese"):
				if not os.path.basename(root).startswith("zh_"):
					continue
				pinyin = os.path.basename(root)
				index = provinces.index(pinyin) + PROVINCE_START + 1 #1是拼音对应的汉字
				for filename in files:
					filepath = os.path.join(root,filename)
					digit_img = cv2.imread(filepath)
					digit_img = cv2.cvtColor(digit_img, cv2.COLOR_BGR2GRAY)
					chars_train.append(digit_img)
					#chars_label.append(1)
					chars_label.append(index)
			chars_train = list(map(deskew, chars_train))
			chars_train = preprocess_hog(chars_train)
			#chars_train = chars_train.reshape(-1, 20, 20).astype(np.float32)
			chars_label = np.array(chars_label)
			print(chars_train.shape)
			self.modelchinese.train(chars_train, chars_label)

	def save_traindata(self):
		if not os.path.exists("svm.dat"):
			self.model.save("svm.dat")
		if not os.path.exists("svmchinese.dat"):
			self.modelchinese.save("svmchinese.dat")

	def accurate_place(self, card_img_hsv, limit1, limit2, color):
		row_num, col_num = card_img_hsv.shape[:2]
		xl = col_num
		xr = 0
		yh = 0
		yl = row_num
		#col_num_limit = self.cfg["col_num_limit"]
		row_num_limit = self.cfg["row_num_limit"]
		col_num_limit = col_num * 0.8 if color != "green" else col_num * 0.5#绿色有渐变
		for i in range(row_num):
			count = 0
			for j in range(col_num):
				H = card_img_hsv.item(i, j, 0)
				S = card_img_hsv.item(i, j, 1)
				V = card_img_hsv.item(i, j, 2)
				if limit1 < H <= limit2 and 34 < S and 46 < V:
					count += 1
			if count > col_num_limit:
				if yl > i:
					yl = i
				if yh < i:
					yh = i
		for j in range(col_num):
			count = 0
			for i in range(row_num):
				H = card_img_hsv.item(i, j, 0)
				S = card_img_hsv.item(i, j, 1)
				V = card_img_hsv.item(i, j, 2)
				if limit1 < H <= limit2 and 34 < S and 46 < V:
					count += 1
			if count > row_num - row_num_limit:
				if xl > j:
					xl = j
				if xr < j:
					xr = j
		return xl, xr, yh, yl
		
	def predict(self, car_pic):
		if type(car_pic) == type(""):
			img = imreadex(car_pic)
		else:
			img = car_pic
		pic_hight, pic_width = img.shape[:2]

		if pic_width > MAX_WIDTH:
			resize_rate = MAX_WIDTH / pic_width
			img = cv2.resize(img, (MAX_WIDTH, int(pic_hight*resize_rate)), interpolation=cv2.INTER_AREA)
		
		blur = self.cfg["blur"]
		#高斯去噪
		if blur > 0:
			img = cv2.GaussianBlur(img, (blur, blur), 0)#图片分辨率调整
		oldimg = img
		img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
		#equ = cv2.equalizeHist(img)
		#img = np.hstack((img, equ))
		#去掉图像中不会是车牌的区域
		kernel = np.ones((20, 20), np.uint8)
		img_opening = cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel)
		img_opening = cv2.addWeighted(img, 1, img_opening, -1, 0);

		#找到图像边缘
		ret, img_thresh = cv2.threshold(img_opening, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
		img_edge = cv2.Canny(img_thresh, 100, 200)
		#使用开运算和闭运算让图像边缘成为一个整体
		kernel = np.ones((self.cfg["morphologyr"], self.cfg["morphologyc"]), np.uint8)
		img_edge1 = cv2.morphologyEx(img_edge, cv2.MORPH_CLOSE, kernel)
		img_edge2 = cv2.morphologyEx(img_edge1, cv2.MORPH_OPEN, kernel)

		#查找图像边缘整体形成的矩形区域，可能有很多，车牌就在其中一个矩形区域中
		try:
			contours, hierarchy = cv2.findContours(img_edge2, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
		except ValueError:
			image, contours, hierarchy = cv2.findContours(img_edge2, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
		contours = [cnt for cnt in contours if cv2.contourArea(cnt) > Min_Area]
		print('len(contours)', len(contours))
		#一一排除不是车牌的矩形区域
		car_contours = []
		for cnt in contours:
			rect = cv2.minAreaRect(cnt)
			area_width, area_height = rect[1]
			if area_width < area_height:
				area_width, area_height = area_height, area_width
			wh_ratio = area_width / area_height
			#print(wh_ratio)
			#要求矩形区域长宽比在2到5.5之间，2到5.5是车牌的长宽比，其余的矩形排除
			if wh_ratio > 2 and wh_ratio < 5.5:
				car_contours.append(rect)
				box = cv2.boxPoints(rect)
				box = np.int0(box)
				#oldimg = cv2.drawContours(oldimg, [box], 0, (0, 0, 255), 2)
				#cv2.imshow("edge4", oldimg)
				#print(rect)

		print(len(car_contours))

		print("精确定位")
		card_imgs = []
		#矩形区域可能是倾斜的矩形，需要矫正，以便使用颜色定位
		for rect in car_contours:
			if rect[2] > -1 and rect[2] < 1:#创造角度，使得左、高、右、低拿到正确的值
				angle = 1
			else:
				angle = rect[2]
			rect = (rect[0], (rect[1][0]+5, rect[1][1]+5), angle)#扩大范围，避免车牌边缘被排除

			box = cv2.boxPoints(rect)
			heigth_point = right_point = [0, 0]
			left_point = low_point = [pic_width, pic_hight]
			for point in box:
				if left_point[0] > point[0]:
					left_point = point
				if low_point[1] > point[1]:
					low_point = point
				if heigth_point[1] < point[1]:
					heigth_point = point
				if right_point[0] < point[0]:
					right_point = point

			if left_point[1] <= right_point[1]:#正角度
				new_right_point = [right_point[0], heigth_point[1]]
				pts2 = np.float32([left_point, heigth_point, new_right_point])#字符只是高度需要改变
				pts1 = np.float32([left_point, heigth_point, right_point])
				M = cv2.getAffineTransform(pts1, pts2)
				dst = cv2.warpAffine(oldimg, M, (pic_width, pic_hight))
				point_limit(new_right_point)
				point_limit(heigth_point)
				point_limit(left_point)
				card_img = dst[int(left_point[1]):int(heigth_point[1]), int(left_point[0]):int(new_right_point[0])]
				card_imgs.append(card_img)
				#cv2.imshow("card", card_img)
				#cv2.waitKey(0)
			elif left_point[1] > right_point[1]:#负角度
				
				new_left_point = [left_point[0], heigth_point[1]]
				pts2 = np.float32([new_left_point, heigth_point, right_point])#字符只是高度需要改变
				pts1 = np.float32([left_point, heigth_point, right_point])
				M = cv2.getAffineTransform(pts1, pts2)
				dst = cv2.warpAffine(oldimg, M, (pic_width, pic_hight))
				point_limit(right_point)
				point_limit(heigth_point)
				point_limit(new_left_point)
				card_img = dst[int(right_point[1]):int(heigth_point[1]), int(new_left_point[0]):int(right_point[0])]
				card_imgs.append(card_img)
				#cv2.imshow("card", card_img)
				#cv2.waitKey(0)
		#开始使用颜色定位，排除不是车牌的矩形，目前只识别蓝、绿、黄车牌
		colors = []
		for card_index,card_img in enumerate(card_imgs):
			green = yello = blue = black = white = 0
			card_img_hsv = cv2.cvtColor(card_img, cv2.COLOR_BGR2HSV)
			#有转换失败的可能，原因来自于上面矫正矩形出错
			if card_img_hsv is None:
				continue
			row_num, col_num= card_img_hsv.shape[:2]
			card_img_count = row_num * col_num

			for i in range(row_num):
				for j in range(col_num):
					H = card_img_hsv.item(i, j, 0)
					S = card_img_hsv.item(i, j, 1)
					V = card_img_hsv.item(i, j, 2)
					if 11 < H <= 34 and S > 34:#图片分辨率调整
						yello += 1
					elif 35 < H <= 99 and S > 34:#图片分辨率调整
						green += 1
					elif 99 < H <= 124 and S > 34:#图片分辨率调整
						blue += 1
					
					if 0 < H <180 and 0 < S < 255 and 0 < V < 46:
						black += 1
					elif 0 < H <180 and 0 < S < 43 and 221 < V < 225:
						white += 1
			color = "no"

			limit1 = limit2 = 0
			if yello*2 >= card_img_count:
				color = "yello"
				limit1 = 11
				limit2 = 34#有的图片有色偏偏绿
			elif green*2 >= card_img_count:
				color = "green"
				limit1 = 35
				limit2 = 99
			elif blue*2 >= card_img_count:
				color = "blue"
				limit1 = 100
				limit2 = 124#有的图片有色偏偏紫
			elif black + white >= card_img_count*0.7:#TODO
				color = "bw"
			print(color)
			colors.append(color)
			print(blue, green, yello, black, white, card_img_count)
			#cv2.imshow("color", card_img)
			#cv2.waitKey(0)
			if limit1 == 0:
				continue
			#以上为确定车牌颜色
			#以下为根据车牌颜色再定位，缩小边缘非车牌边界
			xl, xr, yh, yl = self.accurate_place(card_img_hsv, limit1, limit2, color)
			if yl == yh and xl == xr:
				continue
			need_accurate = False
			if yl >= yh:
				yl = 0
				yh = row_num
				need_accurate = True
			if xl >= xr:
				xl = 0
				xr = col_num
				need_accurate = True
			card_imgs[card_index] = card_img[yl:yh, xl:xr] if color != "green" or yl < (yh-yl)//4 else card_img[yl-(yh-yl)//4:yh, xl:xr]
			if need_accurate:#可能x或y方向未缩小，需要再试一次
				card_img = card_imgs[card_index]
				card_img_hsv = cv2.cvtColor(card_img, cv2.COLOR_BGR2HSV)
				xl, xr, yh, yl = self.accurate_place(card_img_hsv, limit1, limit2, color)
				if yl == yh and xl == xr:
					continue
				if yl >= yh:
					yl = 0
					yh = row_num
				if xl >= xr:
					xl = 0
					xr = col_num
			card_imgs[card_index] = card_img[yl:yh, xl:xr] if color != "green" or yl < (yh-yl)//4 else card_img[yl-(yh-yl)//4:yh, xl:xr]
		#以上为车牌定位
		#以下为识别车牌中的字符
		predict_result = []
		roi = None
		card_color = None
		for i, color in enumerate(colors):
			if color in ("blue", "yello", "green"):
				card_img = card_imgs[i]
				gray_img = cv2.cvtColor(card_img, cv2.COLOR_BGR2GRAY)
				#黄、绿车牌字符比背景暗、与蓝车牌刚好相反，所以黄、绿车牌需要反向
				if color == "green" or color == "yello":
					gray_img = cv2.bitwise_not(gray_img)
				ret, gray_img = cv2.threshold(gray_img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
				#查找水平直方图波峰
				x_histogram  = np.sum(gray_img, axis=1)
				x_min = np.min(x_histogram)
				x_average = np.sum(x_histogram)/x_histogram.shape[0]
				x_threshold = (x_min + x_average)/2
				wave_peaks = find_waves(x_threshold, x_histogram)
				if len(wave_peaks) == 0:
					print("peak less 0:")
					continue
				#认为水平方向，最大的波峰为车牌区域
				wave = max(wave_peaks, key=lambda x:x[1]-x[0])
				gray_img = gray_img[wave[0]:wave[1]]
				#查找垂直直方图波峰
				row_num, col_num= gray_img.shape[:2]
				#去掉车牌上下边缘1个像素，避免白边影响阈值判断
				gray_img = gray_img[1:row_num-1]
				y_histogram = np.sum(gray_img, axis=0)
				y_min = np.min(y_histogram)
				y_average = np.sum(y_histogram)/y_histogram.shape[0]
				y_threshold = (y_min + y_average)/5#U和0要求阈值偏小，否则U和0会被分成两半

				wave_peaks = find_waves(y_threshold, y_histogram)

				#for wave in wave_peaks:
				#	cv2.line(card_img, pt1=(wave[0], 5), pt2=(wave[1], 5), color=(0, 0, 255), thickness=2) 
				#车牌字符数应大于6
				if len(wave_peaks) <= 6:
					print("peak less 1:", len(wave_peaks))
					continue
				
				wave = max(wave_peaks, key=lambda x:x[1]-x[0])
				max_wave_dis = wave[1] - wave[0]
				#判断是否是左侧车牌边缘
				if wave_peaks[0][1] - wave_peaks[0][0] < max_wave_dis/3 and wave_peaks[0][0] == 0:
					wave_peaks.pop(0)
				
				#组合分离汉字
				cur_dis = 0
				for i,wave in enumerate(wave_peaks):
					if wave[1] - wave[0] + cur_dis > max_wave_dis * 0.6:
						break
					else:
						cur_dis += wave[1] - wave[0]
				if i > 0:
					wave = (wave_peaks[0][0], wave_peaks[i][1])
					wave_peaks = wave_peaks[i+1:]
					wave_peaks.insert(0, wave)
				
				#去除车牌上的分隔点
				point = wave_peaks[2]
				if point[1] - point[0] < max_wave_dis/3:
					point_img = gray_img[:,point[0]:point[1]]
					if np.mean(point_img) < 255/5:
						wave_peaks.pop(2)
				
				if len(wave_peaks) <= 6:
					print("peak less 2:", len(wave_peaks))
					continue
				part_cards = seperate_card(gray_img, wave_peaks)
				for i, part_card in enumerate(part_cards):
					#可能是固定车牌的铆钉
					if np.mean(part_card) < 255/5:
						print("a point")
						continue
					part_card_old = part_card
					w = abs(part_card.shape[1] - SZ)//2
					
					part_card = cv2.copyMakeBorder(part_card, 0, 0, w, w, cv2.BORDER_CONSTANT, value = [0,0,0])
					part_card = cv2.resize(part_card, (SZ, SZ), interpolation=cv2.INTER_AREA)
					
					#part_card = deskew(part_card)
					part_card = preprocess_hog([part_card])
					if i == 0:
						resp = self.modelchinese.predict(part_card)
						charactor = provinces[int(resp[0]) - PROVINCE_START]
					else:
						resp = self.model.predict(part_card)
						charactor = chr(resp[0])
					#判断最后一个数是否是车牌边缘，假设车牌边缘被认为是1
					if charactor == "1" and i == len(part_cards)-1:
						if part_card_old.shape[0]/part_card_old.shape[1] >= 7:#1太细，认为是边缘
							continue
					predict_result.append(charactor)
				roi = card_img
				card_color = color
				break
				
		return predict_result, roi, card_color#识别到的字符、定位的车牌图像、车牌颜色

if __name__ == '__main__':
	c = CardPredictor()
	c.train_svm()
	r, roi, color = c.predict("黑A16341.jpg")
	print(r)