About

A computer vision toolkit focused on color detection and feature matching using OpenCV. It allows you to easily start the picamera in case you're using a Raspberry PI.

Some of the stuff you can currently do

• Color detection
  • Detect a range of colors in an image using HSV boundaries.
  • Find bounding boxes.
• Feature matching
  • Draw matches between a source and target image.
  • Find bounding boxes.
• Picamera
  • Easily start the picamera.
• Tools
  • Draw boxes.
  • Draw boxes' offset from the center of the frame.
  • Stack frames in a grid.

Dependencies

Dependency	Installation
python3	Refer to the official website
opencv	Refer to the official installation guide (tested with version 4.5.2)
numpy	pip install numpy
picamera	Installed by default in Raspberry PI OS (required only if working with a picamera)

Installation

pip install cv-recon

Usage

See examples in the examples folder or test it directly form source. Change directory cd cv_recon/recon/ once in this folder you can run:

Command	Description	Preview
python colorspace.py	Generates HSV settings to detect a specific color
python colorspace.py <path to .log file>	Loads HSV settings to detect a specific color
python features.py <path to an image>	Performs feature detection against a given image

Documentation

Class: Colorspace(hsv_settings=None)

This class allows you to detect a range of colors using HSV boundaries. You can generate the settings or set them directly. See examples here.

Args	Description	Default
hsv_settings	Path to .log file containing the HSV settings or list containing lower and upper HSV boundaries	None

Import example

from cv_recon import Colorspace
# load generated settings
colorspace_1 = Colorspace('settings.log')
# or set hsv lower and upper boundaries
colorspace_2 = Colorspace([ [0, 0, 0], [179, 255, 255] ])

Properties

Property	Description	Type	Default
lower	Lower HSV boundary	list	None
upper	Upper HSV boundary	list	None
im_mask	Mask obtained from the HSV boundaries	np.array	None
im_cut	Portions of the frame containing the color boundaries	np.array	None
im_edges	Canny edge detection applied to im_mask	np.array	None
im_contours	Contours of the detected objects drawn on the current frame	np.array	None

Methods

loadSettings(settings)

Loads HSV settings from a generated .log file.

Args	Description	Default
settings	Path to .log file with generated HSV settings	None

returns: None

dumpSettings(output='last.log')

Generates a .log file with the current HSV settings.

Args	Description	Default
output	Path in which the file is gonna be written	'last.log'

returns: None

createSliders()

Creates a window with sliders in order to adjust the HSV settings.
returns: None

updateHSV()

Updates the current HSV settings with the current slider values.
returns: None

getMaskBoxes(im_base, im_hsv, min_area=20, scale=0.1)

Generates a list containing the bounding boxes (x, y, w, h) of the objects.

Args	Description	Default
im_base	Base image in bgr format	None
im_hsv	Base image in hsv format	None
min_area	Minimum area to generate the coordinates	20
scale	Scale of the bounding box	0.1

returns: bounding_boxes

getMaskBoxesArea(im_base, im_hsv, min_area=20, scale=0.1)

Generates two lists containing the bounding boxes (x, y, w, h) and the estimated area of each object.

Args	Description	Default
im_base	Base image in bgr format	None
im_hsv	Base image in hsv format	None
min_area	Minimum area to generate the coordinates	20
scale	Scale of the bounding box	0.1

returns: bounding_boxes, areas

Class: Features(im_source=None, features=500)

This class allows you to easily perform feature matching detection. See examples here.

Args	Description	Default
im_source	Source image	None
features	Amount of features in im_source	500

Import example

from cv_recon import Features
import cv2 as cv

# load source image (the image you want to detect)
im_source = cv.imread('image.jpg')
# create Features object (detects 1000 features from the source image)
my_feature = Features(im_source, 1000)

Properties

Property	Description	Type	Default
im_source	Source image (the image you want to detect)	np.array	im_source
im_source_kp	Source image keypoints	np.array	im_source keypoints
im_target	Target image	np.array	None
im_target_kp	Target image keypoints	np.array	None
im_poly	Image containing a polygon around the best matches	np.array	None

Methods

loadTarget(im)

Loads the target image to perform the feature matching detection.

Args	Description	Default
im	Target image in which the feature matching is gonna be perform	None

returns: None

getMatches(distance=0.75)

Generates a list with the good matches found in the target image.

Args	Description	Default
distance	Threshold which decides if it is a good match	0.75

returns: good_matches

matchPoints(matches)

Returns an image containing the matches between im_target and im_source.

Args	Description	Default
matches	List containing the good matches	None

returns: image

getBoxes(matches, min_matches=20)

Generates a list containing the bounding box (x, y, w, h) of the object.

Args	Description	Default
matches	Good matches	None
min_matches	Minimum amount of matches to generate the bounding box	20

returns: bounding_box

Class: PiCam(resolution=(320, 240), framerate=32, **kargs)

This class allows you to easily interact with the picamera. See examples here.

Args	Description	Default
resolution	Camera resolution	(320, 240)
framerate	Framerate	32
**kargs	Assign default picamera settings. See a list of the settings here	None

Import example

from cv_recon.picam import PiCam

# cam settings
res = (320, 240)
fps = 24

# initialize the camera
camera = PiCam(res, fps, brightness=55, contrast=10)

Properties

Property	Description	Type
current_frame	Current frame	np.array

Methods

videoCapture()

Creates a thread which updates the property current_frame.
returns: None

release()

Stops updating the property current_frame.
returns: None

effects()

Prints the list of image effects.
returns: None

exposureModes()

Prints the list of exposure modes.
returns: None

awbModes()

Prints the list of automatic withe balance modes.
returns: None

Module: cv_tools

This module allows you generate a grid of images, draw bounding boxes and its offset from the center of the frame.

Import example

from cv_recon import cv_tools

Functions

grid(base, dimensions, images, scale=0.5)

Generates a numpy.array containing a grid of images with the given dimensions and scale.

Args	Description	Default
base	Image with the base dimensions for the rest of the images	None
dimensions	Tupla containing the dimensions of the grid	None
images	List of images not larger than dimensions[0] * dimensions[1], each image must have the same dimensions as base	None
scale	Scale of the output image	0.5

returns: image

getBoxesOffset(im, boxes)

Generates a list containing the offset of each box from the center of the frame.

Args	Description	Default
im	Image with the size of the frame	None
boxes	List of bounding boxes	None

returns: [x_offset, y_offset]

drawBoxes(im, boxes)

Draw the bounding boxes over an image.

Args	Description	Default
im	Image in which the bounding boxes are going to be drawn	None
boxes	List of bounding boxes	None

returns: image

drawBoxesPos(im, boxes)

Draw the offset from the center of the frame of each bounding box.

Args	Description	Default
im	Image in which the offsets are going to be drawn	None
boxes	List of bounding boxes	None

returns: image