Then set appropriate values for minCircularity and maxCircularity. To filter by circularity, set filterB圜ircularity = 1.

a regular hexagon has higher circularity than say a square. This just measures how close to a circle the blob is. By Shape : Now shape has three different parameters. setting minArea = 100 will filter out all the blobs that have less then 100 pixels. By Size : You can filter the blobs based on size by setting the parameters filterByArea = 1, and appropriate values for minArea and maxArea. Set blobColor = 0 to select darker blobs, and blobColor = 255 for lighter blobs. By Color : įirst you need to set filterB圜olor = 1. The parameters for SimpleBlobDetector can be set to filter the type of blobs we want. Learn More Filtering Blobs by Color, Size and Shape Center & Radius Calculation : The centers and radii of the new merged blobs are computed and returned.Merging : The centers of the binary blobs in the binary images are computed, and blobs located closer than minDistBetweenBlobs are merged.Grouping : In each binary image, connected white pixels are grouped together.So the first threshold is minThreshold, the second is minThreshold + thresholdStep, the third is minThreshold + 2 x thresholdStep, and so on. These thresholds are incremented by thresholdStep until maxThreshold. Thresholding : Convert the source images to several binary images by thresholding the source image with thresholds starting at minThreshold.Scroll down to know how the parameters are set. The algorithm is controlled by parameters ( shown in bold below ) and has the following steps. SimpleBlobDetector, as the name implies, is based on a rather simple algorithm described below. DrawMatchesFlags::DRAW_RICH_KEYPOINTS flag ensures the size of the circle corresponds to the size of blobĭrawKeypoints( im, keypoints, im_with_keypoints, Scalar(0,0,255), DrawMatchesFlags::DRAW_RICH_KEYPOINTS ) Set up the detector with default parameters. Convert the frame into grayscale grayscale = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)įor many image pre-processing steps, it is ideal to convert it into a grayscale image because it is easier to work with an image of only 1 channel (without the need to worry about 3 channels in colored images) and the task at hand does not require the image to be colored since lanes can still be observed in grayscale images, saving a lot of computation time and resources.Mat im = imread( 'blob.jpg', IMREAD_GRAYSCALE ) While these steps are not new because there have been countless experiments on how best to pre-process each frame before detecting lines, I try to use my past experiences on image pre-processing to give it a slightly different result and explain the reasons behind doing so at the same time. Now, let’s get into some image pre-processing with only a few lines of code. This saves us some work because then we only need to pay attention to the bottom part of the frame for preprocessing and computation. The camera is fixed on the car dashboard, hence the road lanes are visible only within the bottom part of the frame.

Before we get into detecting lanes, we have to set some “house rules”: In a future article, I’ll try to implement deep learning techniques for lane detection before integrating it with a Raspberry Pi for our wheeled robot. This article focuses on lane detection on highway roads using OpenCV as an example. The navigation system of the robot depended entirely on the infrared sensors guiding its path forward, which does the job, but not very good at adjusting to dynamic conditions. A very big part of lane detection using OpenCV involves Hough Lines Transform, hence I’ll do my part in explaining Hough Lines Transform as best as I can. This week, I tried to implement lane detection using OpenCV to improve its navigation ability. Its navigation depended entirely on the infrared sensors guiding its path forward.

In one of my previous articles, I mentioned we built a meal delivery robot for our capstone project integrating some simple OpenCV and OCR tricks to enable the robot to detect patients.

Limb detection opencv