1

I am trying to iterate the source array in order to change a few elements in the destination array, but I can’t get the correct indexes, I have incomprehensible offsets. In general, my task is to change the element if the conditions require it 

import numpy as np

x = np.linspace(0,1000, 1000/50)
y = np.linspace(0,1000, 1000/50)
X,Y = np.meshgrid(x,y)

source =  np.column_stack([X.ravel(), Y.ravel()]).astype(int)
destination = source.copy()

for x, i in np.ndenumerate(source):
    # if smth:
    destination[i] = np.array([x[0] + 10, x[1]])

I think I should not use the source array, but only iterate over the destination array(сannot be done using standard methods), tell me the correct solution, thanks in advance.

Current output:
[421 789]
[473 789]
[526 789]
[578 789]
[631 789]
[684 789]


Required output:
[431 789]
[483 789]
[536 789]
[588 789]
[641 789]
[694 789]

I’ll explain more simply, I have a grid, it has points, I need to shift the points, say 88, 89, 90, 10 pixels to the right, for this I need to have an array of source and destination (where these points are offset), enumerate most likely it doesn’t suit me, but the usual editing of an array like for x in destination: when editing x gives the desired result, but this does not apply to ndarray

enter image description here

for x, i in enumerate(destination):
     inside = cv2.pointPolygonTest(cnt, (destination[x,0], 
     destination[x,1]), False)
   if inside > 0:
    cv2.circle(img, (destination[x,0], destination[x,1]), 10, 
  (255,0,0), 2)
    destination[x] = np.array([destination[x,0] + 10, destination[x,1]])

 # Contour(cnt)
 [[550  42]
 [600  42]
 [690 273]
 [640 273]]

As you understand, I need to shift everything that is circled in blue by 10 pixels

Improve this question
8
  • what is the desired output? Commented Oct 7, 2019 at 20:05
  • 1
    if smth: sounds like a task for np.where Commented Oct 7, 2019 at 20:08
  • @QuangHoang, This is a condition with OpenCV, I did not insert it here, maybe someone may not have it Commented Oct 7, 2019 at 20:10
  • how do you get these desired values? what is the mathematical equation behind? Commented Oct 7, 2019 at 20:13
  • @makis I use cv2.pointPolygonTest to check if these points are in the given area, if so then I shift them 10 pixels to the right Commented Oct 7, 2019 at 20:25

2 Answers 2

Reset to default
1

Solution

From data creation to target-region modification, we can divide it into a three step process. The modification can be achieved using numpy indexing along with conditional region selection by means of numpy.where and numpy.logical_and (if necessary).

1. Make Data

import numpy as np

x = np.linspace(0,1000, int(1000/50))
y = np.linspace(0,1000, int(1000/50))
X,Y = np.meshgrid(x,y)

source =  np.column_stack([X.ravel(), Y.ravel()]).astype(int)
destination = source.copy()

2. Use Conditional Statement to Find Target Region

target_index = np.where(np.logical_and(destination[:,1]==789, destination[:,0]>=421))
destination[target_index]

Output: 

array([[ 421,  789],
       [ 473,  789],
       [ 526,  789],
       [ 578,  789],
       [ 631,  789],
       [ 684,  789],
       [ 736,  789],
       [ 789,  789],
       [ 842,  789],
       [ 894,  789],
       [ 947,  789],
       [1000,  789]])

3. Make changes to the target region

scope = destination[target_index] 
scope[:,0] = scope[:,0] + 10
destination[target_index] = scope
destination[target_index]

Output: 

array([[ 431,  789],
       [ 483,  789],
       [ 536,  789],
       [ 588,  789],
       [ 641,  789],
       [ 694,  789],
       [ 746,  789],
       [ 799,  789],
       [ 852,  789],
       [ 904,  789],
       [ 957,  789],
       [1010,  789]])
Improve this answer
Sign up to request clarification or add additional context in comments.

4 Comments

Thanks, the solution is very close, but I need to change the points in a certain shape. I don't know how to use np where for this. I will update the description, thanks in advance.
@question_mark_77 I took a look at the image and edited it to focus on the example you have shown. What I understand is that you have an objective that can be broken down into multiple smaller objectives. This answer served the purpose of explaining the question you asked. However, the problem you are posing now, is a different problem: you need to find out your business logic to determine those points and their offsets. Please create a separate Stackoverflow question with a tag to imageprocessing. Leave a reference to this question there. Please mark an answers as the accepted one.
Not a problem, just one theoretical question, can I use np.where for these purposes?
Theoretically speaking, yes. np.where basically means where(some condition). So, as long as you have your condition defined properly and that evaluates to an array of Boolean values, you should be able to use np.where.
1

What you are asking is not clear at all. How to do you want to modify the destination?

Also, are you sure you need np.ndenumerate and not enumerate ?

Here is a way to modify it based on the values that are in sources. At the position (index) x, the destination will become equal to the x element of the first column of source+ 10 and equal to the x element of the second column of source.

for index, value in enumerate(source):
    destination[index] = [source[index,0]+10, source[index,1]] # x+10, y same
Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.