CODEWITHSUNDEEP
androidopencv
P.Oni
P.Oni

Reputation: 45

Why I'm Getting wrong values with Colour detection (OpenCV)

I have a problem with conversion from BGR to HSV. I'm programming with Android Studio and testing with my Xperia Z5.

In my code snippet, I'm getting totally wrong colour values:

Scalar LOWER_RED  = (0,0,0);
Scalar HIGHER_RED = (30,255,255);

Mat src = new Mat(Bitmap.getHeight(), Bitmap.getWidth(),CvType.CV_8UC4);
Mat hsv = new Mat(Bitmap.getHeight(), Bitmap.getWidth(),CvType.CV_8UC4);

Utils bitmapToMat(Bitmap, src);

Imgproc.cvtColor(src,hsv,Imgproc.COLOR_BGR2HSV);

Core.inRange(hsv, LOWER_RED, HIGHER_RED, hsv);

Utils.matToBitmap(hsv,Bitmap);

I want to capture red colour. What did I do wrong?

Edit: I tried with all advices and my Code Snippet looks now this way: 

Scalar LOWER_RED  = (0,10,100);
Scalar HIGHER_RED = (10,255,255);

Mat src = new Mat(Bitmap.getHeight(), Bitmap.getWidth(),CvType.CV_8UC3);
Mat hsv = new Mat(Bitmap.getHeight(), Bitmap.getWidth(),CvType.CV_8UC3);

Utils bitmapToMat(Bitmap, src);

Imgproc.cvtColor(src,hsv,Imgproc.COLOR_BGR2HSV);

Core.inRange(hsv, LOWER_RED, HIGHER_RED, hsv);

Utils.matToBitmap(hsv,Bitmap);

The Outcome is a black screen ( no matches )

with 

Core.inRange(hsv,New Scalar(0,0,0),New Scalar(10,255,255),HighRedRange);
Core.inRange(hsv,New Scalar(160,100,100),New Scalar(179,255,255),LowRedRange);
Core.addWeighted(LowRedRange,1.0,HighredRange,1.0,0.0,hsv);

The vegetables are black and the white background is white in hsv 0,0,0 - 10,255,255 AND 160,100,100 - 179,255,255

If I use a Scalar from 110,100,100 until 135,255,255, then the red pepper is white and the back ground black ( correctly detected ).

Source Picture:

enter image description here

And I dont understand all this...

Upvotes: 2

Views: 1468

Answers (2)

P.Oni
P.Oni

Reputation: 45

I know now my problem it's this:

Imgproc.cvtColor(src,hsv,Imgproc.COLOR_BGR2HSV);

With RGB2HSV all values are correct. I thought on Android Smartphones there is BGR used ?

However, Big thanks for all answers.

I wish all of you a great day :)

Upvotes: 1

cagatayodabasi
cagatayodabasi

Reputation: 762

There is a good tutorial here. It's for C++ but the general idea is the same. I tried the general idea and it surely works. The problem is that your range is too broad. In OpenCV, Hue range is in 0-180. Meaning that your higher limit goes to 30*2 = 60 which includes nearly all yellow range too.

I set the range from 0 to 10 for Hue, but remember you may also want to get 160 - 179 range which also includes some part of red. For this, you just need a second mask and then combine them with simple addition.

The example code in Python:

import cv2

import numpy as np

img = cv2.imread('peppers.jpg',1)

im_hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)

thresh_low = np.array([0,100,100])
thresh_high = np.array([10,255,255]) 

mask = cv2.inRange(im_hsv, thresh_low, thresh_high)

im_masked = cv2.bitwise_and(img,img, mask= mask)

cv2.imshow('Masked',im_masked)

cv2.waitKey(0)

Original image:

enter image description here

Result:

enter image description here

Upvotes: 2

Related Questions