基于opencv的图像阴影消除&车辆变道检测

基于opencv的图像阴影消除

详细代码在这！！！

最大滤波

def max_filtering(N, I_temp):
    wall = np.full((I_temp.shape[0]+(N//2)*2, I_temp.shape[1]+(N//2)*2), -1)
    wall[(N//2):wall.shape[0]-(N//2), (N//2):wall.shape[1]-(N//2)] = I_temp.copy()
    temp = np.full((I_temp.shape[0]+(N//2)*2, I_temp.shape[1]+(N//2)*2), -1)
    for y in range(0,wall.shape[0]):
        for x in range(0,wall.shape[1]):
            if wall[y,x]!=-1:
                window = wall[y-(N//2):y+(N//2)+1,x-(N//2):x+(N//2)+1]
                num = np.amax(window)
                temp[y,x] = num
    A = temp[(N//2):wall.shape[0]-(N//2), (N//2):wall.shape[1]-(N//2)].copy()
    return A

1. np.full：返回一个指定形状、类型和数值的数组

numpy.full(shape, fill_value, dtype=None, order=‘C’)

• shape
• fill_value

2. 前三步图示一下

 wall = np.full((I_temp.shape[0]+(N//2)*2, I_temp.shape[1]+(N//2)*2), -1)

+(N//2)*2有点类似于padding

最小滤波

def min_filtering(N, A):
    wall_min = np.full((A.shape[0]+(N//2)*2, A.shape[1]+(N//2)*2), 300)
    wall_min[(N//2):wall_min.shape[0]-(N//2), (N//2):wall_min.shape[1]-(N//2)] = A.copy()
    temp_min = np.full((A.shape[0]+(N//2)*2, A.shape[1]+(N//2)*2), 300)
    for y in range(0,wall_min.shape[0]):
        for x in range(0,wall_min.shape[1]):
            if wall_min[y,x]!=300:
                window_min = wall_min[y-(N//2):y+(N//2)+1,x-(N//2):x+(N//2)+1]
                num_min = np.amin(window_min)
                temp_min[y,x] = num_min
    B = temp_min[(N//2):wall_min.shape[0]-(N//2), (N//2):wall_min.shape[1]-(N//2)].copy()
    return B

def background_subtraction(I, B):
O = I - B
norm_img = cv2.normalize(O, None, 0,255, norm_type=cv2.NORM_MINMAX)
return norm_img

最小滤波和最大滤波的顺序：

①
②

def min_max_filtering(M, N, I):
    if M == 0:
        #max_filtering
        A = max_filtering(N, I)
        #min_filtering
        B = min_filtering(N, A)
        #subtraction
        normalised_img = background_subtraction(I, B)
    elif M == 1:
        #min_filtering
        A = min_filtering(N, I)
        #max_filtering
        B = max_filtering(N, A)
        #subtraction
        normalised_img = background_subtraction(I, B)
    return normalised_img

基于opencv的车辆变道检测

代码在基于opencv的车辆变道检测
整体结构如下：

angle.py解析

找到汽车的角度

import cv2
import math


class Target:
    def __init__(self):
        self.capture = cv2.VideoCapture(0)
        cv2.namedWindow("Target", 1)
        cv2.NamedWindow("Threshold1",1)
        cv2.NamedWindow("Threshold2",1)
        cv2.NamedWindow("hsv",1)
     
    def run(self):
        #initiate font
        font = cv2.InitFont(cv2.CV_FONT_HERSHEY_SIMPLEX, 1, 1, 0, 3, 8)
        #instantiate images
        hsv_img=cv2.CreateImage(cv2.GetSize(cv2.QueryFrame(self.capture)),8,3)
        threshold_img1 = cv2.CreateImage(cv2.GetSize(hsv_img),8,1)
        threshold_img1a = cv2.CreateImage(cv2.GetSize(hsv_img),8,1)
        threshold_img2 = cv2.CreateImage(cv2.GetSize(hsv_img),8,1)
        i=0
        writer=cv2.CreateVideoWriter("angle_tracking.avi",cv2.CV_FOURCC('M','J','P','G'),30,cv2.GetSize(hsv_img),1)

        while True:
            #capture the image from the cam
            img=cv2.QueryFrame(self.capture)

            #convert the image to HSV
            cv2.CvtColor(img,hsv_img,cv2.CV_BGR2HSV)

            #threshold the image to isolate two colors
            cv2.InRangeS(hsv_img,(165,145,100),(250,210,160),threshold_img1)  #red
            cv2.InRangeS(hsv_img,(0,145,100),(10,210,160),threshold_img1a)   #red again
            cv2.Add(threshold_img1,threshold_img1a,threshold_img1)               #this is combining the two limits for red
            cv2.InRangeS(hsv_img,(105,180,40),(120,260,100),threshold_img2) #blue


            #determine the moments of the two objects
            threshold_img1=cv2.GetMat(threshold_img1)
            threshold_img2=cv2.GetMat(threshold_img2)
            moments1=cv2.Moments(threshold_img1,0)
            moments2=cv2.Moments(threshold_img2,0)
            area1=cv2.GetCentralMoment(moments1,0,0)
            area2=cv2.GetCentralMoment(moments2,0,0)
             
            #initialize x and y
            x1,y1,x2,y2=(1,2,3,4)
            coord_list=[x1,y1,x2,y2]
            for x in coord_list:
                x=0
             
            #there can be noise in the video so ignore objects with small areas
            if (area1 >200000):
                #x and y coordinates of the center of the object is found by dividing the 1,0 and 0,1 moments by the area
                x1=int(cv2.GetSpatialMoment(moments1,1,0)/area1)
                y1=int(cv2.GetSpatialMoment(moments1,0,1)/area1)

                #draw circle
                cv2.Circle(img,(x1,y1),2,(0,255,0),20)

                #write x and y position
                cv2.PutText(img,str(x1)+","+str(y1),(x1,y1+20),font, 255) #Draw the text

            if (area2 >100000):
                #x and y coordinates of the center of the object is found by dividing the 1,0 and 0,1 moments by the area
                x2=int(cv2.GetSpatialMoment(moments2,1,0)/area2)
                y2=int(cv2.GetSpatialMoment(moments2,0,1)/area2)

                #draw circle
                cv2.Circle(img,(x2,y2),2,(0,255,0),20)

                cv2.PutText(img,str(x2)+","+str(y2),(x2,y2+20),font, 255) #Draw the text
                cv2.Line(img,(x1,y1),(x2,y2),(0,255,0),4,cv2.CV_AA)
                #draw line and angle
                cv2.Line(img,(x1,y1),(cv2.GetSize(img)[0],y1),(100,100,100,100),4,cv2.CV_AA)
            x1=float(x1)
            y1=float(y1)
            x2=float(x2)
            y2=float(y2)
            angle = int(math.atan((y1-y2)/(x2-x1))*180/math.pi)
            cv2.PutText(img,str(angle),(int(x1)+50,(int(y2)+int(y1))/2),font,255)

            #cv2.WriteFrame(writer,img)

          
            #display frames to users
            cv2.ShowImage("Target",img)
            cv2.ShowImage("Threshold1",threshold_img1)
            cv2.ShowImage("Threshold2",threshold_img2)
            cv2.ShowImage("hsv",hsv_img)
            # Listen for ESC or ENTER key
            c = cv2.WaitKey(7) % 0x100
            if c == 27 or c == 10:
                break
        cv2.DestroyAllWindows()
             
if __name__=="__main__":
    t = Target()
    t.run()

CreateImageHeader(size, depth, channels)
创建图像标头并分配图像数据

QueryFrame（capture）
在一次调用中结合了 VideoCapture：：grab（） 和 VideoCapture：：retrieve（）。这是读取视频文件或从解码中捕获数据并返回刚刚抓取的帧的最方便的方法。如果未抓取任何帧（相机已断开连接，或者视频文件中没有更多帧），则方法返回 false，函数返回 NULL 指针。

---

将红色和蓝色单独分离开，红色在hsv中有两段，最终做一个加和

#threshold the image to isolate two colors
            cv2.InRangeS(hsv_img,(165,145,100),(250,210,160),threshold_img1)  #red
            cv2.InRangeS(hsv_img,(0,145,100),(10,210,160),threshold_img1a)   #red again
            cv2.Add(threshold_img1,threshold_img1a,threshold_img1)               #this is combining the two limits for red
            cv2.InRangeS(hsv_img,(105,180,40),(120,260,100),threshold_img2) #blue

cv.InRangeS（src， lower， upper， dst）
实际基本等价于InRange

---

 #determine the moments of the two objects
            threshold_img1=cv2.GetMat(threshold_img1)
            threshold_img2=cv2.GetMat(threshold_img2)
            moments1=cv2.Moments(threshold_img1,0)
            moments2=cv2.Moments(threshold_img2,0)
            area1=cv2.GetCentralMoment(moments1,0,0)
            area2=cv2.GetCentralMoment(moments2,0,0)

• GetMat该函数返回输入数组的矩阵标头
• Moments概念:轮廓矩代表了一条轮廓，一幅图像，一组点集（取自《learning opencv》的介绍）
  当处理轮廓时，结果是轮廓的长度；当处理点集时，它代表的是点的数量

cv.Moments（arr， binary=0）

binary如果为 true，则所有非零图像像素都被视为 1。该参数仅用于图像。

GetCentralMoment
GetSpatialMoment

现在好像已经被淘汰，在文档已经找不到了， x2=int(cv2.GetSpatialMoment(moments2,1,0)/area2) y2=int(cv2.GetSpatialMoment(moments2,0,1)/area2)
功能是：通过将1,0和0,1矩除以面积，可以找到对象中心的x和y坐标

还是举一个例子说明不同版本下的使用:

//cv1：
moments = cv.Moments(thresholded_img, 0) 
area = cv.GetCentralMoment(moments, 0, 0) 

#there can be noise in the video so ignore objects with small areas 
if(area > 100000): 
    #determine the x and y coordinates of the center of the object 
    #we are tracking by dividing the 1, 0 and 0, 1 moments by the area 
    x = cv.GetSpatialMoment(moments, 1, 0)/area 
    y = cv.GetSpatialMoment(moments, 0, 1)/area 

//cv2
moments = cv2.moments(thresholded_img) 
area = moments['m00'] 

#there can be noise in the video so ignore objects with small areas 
if(area > 100000): 
    #determine the x and y coordinates of the center of the object 
    #we are tracking by dividing the 1, 0 and 0, 1 moments by the area 
    x = moments['m10'] / area
    y = moments['m01'] / area

GetCentralMoment，GetSpatialMoment在opencv2可以直接使用索引进行获取

angle_pop.py详解

完整代码

import cv2
import numpy as np
import time
import tkinter as tk
from tkinter import ttk

NORM_FONT= ("Verdana", 10)

def popupmsg(msg):
    popup = tk.Tk()
    popup.wm_title("Message")
    label = ttk.Label(popup, text=msg, font=NORM_FONT)
    label.pack(side="top", fill="x", pady=10)
    B1 = ttk.Button(popup, text="Okay", command = popup.destroy)
    B1.pack()
    popup.mainloop()

cascade_src = 'cascade/cars.xml'
video_src = 'dataset/cars.mp4'

cap = cv2.VideoCapture(video_src)
car_cascade = cv2.CascadeClassifier(cascade_src)

deltatsum = 0
n = 0
last_time = time.time()

while(1):
    # Take each frame
    _, frame = cap.read()
    cars = car_cascade.detectMultiScale(frame, 1.1, 1)
    for (x,y,w,h) in cars:
        roi = cv2.rectangle(frame,(x,y),(x+w,y+h),(0,0,255),2)   #ROI is region of interest
    #blur the frame to get rid of noise. the kernel should be ODD
    frame = cv2.GaussianBlur(frame,(21,21),0)
    
    # Convert BGR to HSV
    hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)

    # define range of color in HSV
    lower_limit = np.array([0,150,150])
    upper_limit = np.array([10,255,255])

    # Threshold the HSV image to get only the thresholded colors
    # mask is a binary image
    mask = cv2.inRange(hsv, lower_limit, upper_limit)

    # The dimensions of the kernel must be odd!
    kernel = np.ones((3,3),np.uint8)
    kernel_lg = np.ones((15,15),np.uint8)
    
    # erode the mask to get rid of noise
    mask = cv2.erode(mask,kernel,iterations = 1)

    # dialate it back to regain some lost area
    mask = cv2.dilate(mask,kernel_lg,iterations = 1)    
    
    # Bitwise-AND the mask and grayscale image so we end up with our areas of interest and black everywhere else
    result = cv2.bitwise_and(frame,frame, mask= mask)
    
    thresh = mask
    contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
    
    # define a minimum area for a contour - if it's below this, ignore it 
    min_area = 1000
    cont_filtered = []    
    
    # filter out all contours below a min_area
    for cont in contours:
        if cv2.contourArea(cont) > min_area:
            cont_filtered.append(cont)
            #print(cv2.contourArea(cont))
    try:
        cnt = cont_filtered[0]
        
        # draw the rectangle surrounding the filtered contour
        rect = cv2.minAreaRect(cnt)
        box = cv2.boxPoints(rect)
        box = np.int0(box)
        cv2.drawContours(frame,[box],0,(0,0,255),2)
        
        rows,cols = thresh.shape[:2]
        [vx,vy,x,y] = cv2.fitLine(cnt, cv2.DIST_L2,0,0.01,0.01)
        lefty = int((-x*vy/vx) + y)
        righty = int(((cols-x)*vy/vx)+y)
        cv2.line(frame,(cols-1,righty),(0,lefty),(0,255,0),2)        
        
        # this would draw all the contours on the image, not just the ones from cont_filtered
        #cv2.drawContours(frame, cont_filtered, -1, (0,255,0), 3)

        M = cv2.moments(cnt)
        cx = int(M['m10']/M['m00'])
        cy = int(M['m01']/M['m00'])
        (x,y),(MA,ma),angle = cv2.fitEllipse(cnt)
        print('x= ', cx, '  y= ', cy, ' angle = ', round(rect[2],2))
        if(round(rect[2],2))<-45:
            popupmsg('Lane change detected')
        #print(contours)
        #print('there are contours')
    except:
        print('no contours')

    cv2.imshow('frame',frame)
    #cv2.imshow('mask', mask)
    #cv2.imshow('thresh',thresh)
    #cv2.imshow('im2', im2)
    cv2.imshow('result', result)
    
    k = cv2.waitKey(5) & 0xFF
    if k == ord('q'):
        break
		
    deltat = time.time() - last_time
    last_time = time.time()    
    deltatsum += deltat
    n += 1
    freq = round(1/(deltatsum/n), 2)
    # print('Updating at ' + str(freq) + ' FPS\r', end='')

# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()

模块函数,类解析

import tkinter as tk
from tkinter import ttk

使用了tkinter作为经典ui设计，详细介绍！！！

1.CascadeClassifier(级联分类器)

2.detectMultiScale函数
有多个重载形式，且c++和python传入参数形式有一定区别
输入为灰度形式

整体思想

1. 获取图像,实例化级联分类器
2. 搜索图像，预处理（滤波，阈值…）
3. 找轮廓，筛选轮廓