使用python opencv透视变换getperspectiveTransform获得图像的鸟瞰图

  如何使用透视变换来获得图像的roi的自顶向下的鸟瞰图，首先要确定的是需要变换的参考点。
  建立项目perTran.py， 代码如下：

#! /usr/bin/env python
# -*- coding: utf-8 -*-

import numpy as np
import cv2
import imutils

"""
x小y小rect[0]  x大y小rect[1]       
	-------------
	|           |
	|           |
	-------------
x小y大rect[3]  x大y大rect[2]
"""
def order_points(pts):
	# 初始化矩形4个顶点的坐标
	rect = np.zeros((4, 2), dtype='float32')
	# 坐标点求和 x+y
	s = pts.sum(axis = 1)
	# np.argmin(s) 返回最小值在s中的序号
	rect[0] = pts[np.argmin(s)]
	rect[2] = pts[np.argmax(s)]
	# diff就是后一个元素减去前一个元素  y-x
	diff = np.diff(pts, axis=1)
	rect[1] = pts[np.argmin(diff)]
	rect[3] = pts[np.argmax(diff)]
	# 返回矩形有序的4个坐标点
	return rect

def perTran(image, pts):
	rect = order_points(pts)
	tl, tr, br, bl = rect
	# 计算宽度
	widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
	widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
	maxWidth = max(int(widthA), int(widthB))
	# 计算高度
	heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
	heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
	maxHeight = max(int(heightA), int(heightB))
	# 定义变换后新图像的尺寸
	dst = np.array([[0, 0], [maxWidth-1, 0], [maxWidth-1, maxHeight-1],
				   [0, maxHeight-1]], dtype='float32')
	# 变换矩阵
	M = cv2.getPerspectiveTransform(rect, dst)
	# 透视变换
	warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
	return warped


def main():
	image = cv2.imread('./1.png')
	output = image.copy()
	# 转换成灰度图像
	gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
	# 双边滤波器能够做到平滑去噪的同时还能够很好的保存边缘
	gray = cv2.bilateralFilter(gray, 11, 17, 17)
	# 检测边缘
	edged = cv2.Canny(gray, 30, 200)
	#cv2.imshow('Canny', edged)
	# 查找轮廓
	cnts = cv2.findContours(edged.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
	cnts = imutils.grab_contours(cnts)
	# 获取前3个最大的轮廓
	cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:3]

	screenCnt = None
	for c in cnts:
		# 轮廓周长
		peri = cv2.arcLength(c, True)
		print('arcLength : {:.3f}'.format(peri))
		# approxPolyDP主要功能是把一个连续光滑曲线折线化，对图像轮廓点进行多边形拟合。
		# 近似轮廓的多边形曲线， 近似精度为轮廓周长的1.5%
		approx = cv2.approxPolyDP(c, 0.015 * peri, True)
		# 矩形边框具有4个点， 将其他的剔除
		if len(approx) == 4:
			screenCnt = approx
			break
	# 绘制轮廓矩形边框
	cv2.drawContours(image, [screenCnt], -1, (0, 255, 0), 3)
	# 调整为x,y 坐标点矩阵
	pts = screenCnt.reshape(4, 2)
	#print('screenCnt.reshape:\n{}'.format(pts))
	# 透视变换
	warped = perTran(output, pts)

	cv2.imshow('image', image)
	cv2.imshow('output', output)
	cv2.imshow('warped', warped)

	cv2.waitKey(0)
	cv2.destroyAllWindows()

if __name__ == "__main__":
	main()

  执行代码，结果如下：

总结

  使用opencv的cv2.getPerspctiveTransform函数将图像进行透视变化。最关键的在于确定矩形的顶点，逆时针或顺时针排列都可以，同时确定目标参考点也同样重要。

参考资料

https://www.pyimagesearch.com