Package 'opencv'

OpenCV Computer Vision

Description

Tools to experiment with computer vision algorithms. Use ocv_read and ocv_write to load/save images on disk, or use ocv_picture / ocv_video to use your webcam. In RSudio IDE the image objects will automatically be displayed in the viewer pane.

Usage

ocv_face(image)

ocv_facemask(image)

ocv_read(path)

ocv_write(image, path)

ocv_destroy(image)

ocv_bitmap(image)

ocv_edges(image)

ocv_picture()

ocv_resize(image, width = 0, height = 0)

ocv_mog2(image)

ocv_knn(image)

ocv_hog(image)

ocv_blur(image, ksize = 5)

ocv_sketch(image, color = TRUE)

ocv_stylize(image)

ocv_markers(image)

ocv_info(image)

ocv_copyto(image, target, mask)

ocv_display(image)

ocv_video(filter, stop_on_result = FALSE)

ocv_grayscale(image)

ocv_version()

Arguments

image	an ocv image object created from e.g. ocv_read()
path	image file such as png or jpeg
width	output width in pixels
height	output height in pixels
ksize	size of blurring matrix
color	true or false
target	the output image
mask	only copy pixels from the mask
filter	an R function that takes and returns an opecv image
stop_on_result	stop if an object is detected

Examples

# Silly example
mona <- ocv_read('https://jeroen.github.io/images/monalisa.jpg')

# Edge detection
ocv_edges(mona)
ocv_markers(mona)

# Find face
faces <- ocv_face(mona)

# To show locations of faces
facemask <- ocv_facemask(mona)
attr(facemask, 'faces')

# This is not strictly needed
ocv_destroy(mona)

---

OpenCV keypoints

Description

Find key points in images

Usage

ocv_keypoints(
  image,
  method = c("FAST", "Harris"),
  control = ocv_keypoints_options(method, ...),
  ...
)

Arguments

image	an ocv grayscale image object
method	the type of keypoint detection algorithm
control	a list of arguments passed on to the algorithm
...	further arguments passed on to ocv_keypoints_options

FAST algorithm arguments

• threshold threshold on difference between intensity of the central pixel and pixels of a circle around this pixel.

• nonmaxSuppression if true, non-maximum suppression is applied to detected corners (keypoints).

• type one of the three neighborhoods as defined in the paper: TYPE_9_16, TYPE_7_12, TYPE_5_8

Harris algorithm arguments

• numOctaves the number of octaves in the scale-space pyramid

• corn_thresh the threshold for the Harris cornerness measure

• DOG_thresh the threshold for the Difference-of-Gaussians scale selection

• maxCorners the maximum number of corners to consider

• num_layers the number of intermediate scales per octave

Examples

mona <- ocv_read('https://jeroen.github.io/images/monalisa.jpg')
mona <- ocv_resize(mona, width = 320, height = 477)

# FAST-9
pts <- ocv_keypoints(mona, method = "FAST", type = "TYPE_9_16", threshold = 40)
# Harris
pts <- ocv_keypoints(mona, method = "Harris", maxCorners = 50)

# Convex Hull of points
pts <- ocv_chull(pts)

---

Detect and Decode a QR code

Description

Detect and decode a QR code from an image or camera. By default it returns the text value from the QR code if detected, or NULL if no QR was found. If draw = TRUE then it returns an annotated image with the position and value of the QR drawn into the image, and qr text value as an attribute. The qr_scanner function opens the camera device (if available on your computer) and repeats ocv_qr_detect until it a QR is detected.

Usage

ocv_qr_detect(image, draw = FALSE, decoder = c("wechat", "quirc"))

qr_scanner(draw = FALSE, decoder = c("wechat", "quirc"))

Arguments

image	an ocv image object created from e.g. ocv_read()
draw	if TRUE, the function returns an annotated image showing the position and value of the QR code.
decoder	which decoder implementation to use, see details.

Details

OpenCV has two separate QR decoders. The 'wechat' decoder was added in libopencv 4.5.2 and generally has better performance and fault-tolerance. The old 'quirc' decoder is available on some older versions of libopencv as a plug-in, but many Linux distros did not include it. If you get an error Library QUIRC is not linked. No decoding is performed. this sadly means your Linux distribution is too old and does not support QR decoding.

Value

if a QR code is detected, this returns either the text value of the QR, or if draw it returns the annotated image, with the value as an attribute. Returns NULL if no QR was found in the image.

Examples

png("test.png")
plot(qrcode::qr_code("This is a test"))
dev.off()
ocv_qr_detect(ocv_read('test.png'))
unlink("test.png")

---

OpenCV area manipulation

Description

Manipulate image regions

Usage

ocv_rectangle(image, x = 0L, y = 0L, width, height)

ocv_polygon(image, pts, convex = FALSE, crop = FALSE, color = 255)

ocv_bbox(image, pts)

ocv_chull(pts)

Arguments

image	an ocv image object
x	horizontal location
y	vertical location
width	width of the area
height	height of the area
pts	a list of points with elements x and y
convex	are the points convex
crop	crop the resulting area to its bounding box
color	color for the non-polygon area

Examples

mona <- ocv_read('https://jeroen.github.io/images/monalisa.jpg')

# Rectangular area
ocv_rectangle(mona, x = 400, y = 300, height = 300, width = 350)
ocv_rectangle(mona, x = 0, y = 100, height = 200)
ocv_rectangle(mona, x = 500, y = 0, width = 75)

# Polygon area
img <- ocv_resize(mona, width = 320, height = 477)
pts <- list(x = c(184, 172, 146, 114,  90,  76,  92, 163, 258),
            y = c(72,   68,  70,  90, 110, 398, 412, 385, 210))
ocv_polygon(img, pts)
ocv_polygon(img, pts, crop = TRUE)
ocv_polygon(img, pts, convex = TRUE, crop = TRUE)

# Bounding box based on points
ocv_bbox(img, pts)

# Bounding box of non-zero pixel area
area <- ocv_polygon(img, pts, color = 0, crop = FALSE)
area
area <- ocv_bbox(area)
area

---