To do this well, you will first have to segment the individual blobs in the image. Blob detection is typically adjusted for specific problem domains. I recommend that you check the articles related to your field (I am sure you can find some on Google Scholar or here on RG).

There is no one perfect method for blob detection that will fit all the problems. If you give some more background to your problem domain people will be able to help you more.

You could use cv2.findContours() which results in an array for each contour. Then you can lay a box around it box = cv2.boxPoints(cv2.minAreaRect(c)) and calculate the euclidean mid.

But you need to consider that you have to split joined Objects and the mid of the object isn't necessary the mid of the area in the picture.

Have you tried canny edge detection with Gaussian blur and an automatic threshold ?

https://www.pyimagesearch.com/2015/04/06/zero-parameter-automatic-canny-edge-detection-with-python-and-opencv/

Then apply cv2.findContours() etc. on the resulting edges.

Stefan Schreiber , Thanks for the reply, I will work on trying this method. The end goal is to identify a point on each object for pickup. It doesn't require that the point be the exact center.

Michael Marszalek , Thank you for the reply. I began to write a program to do just that, the results are looking promising thus far. Having an error:

contours, hierarchy = cv2.findContours(edged, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)

ValueError: too many values to unpack

Which CV version are you using ? have a look here:

https://stackoverflow.com/questions/25504964/opencv-python-valueerror-too-many-values-to-unpack

Maybe it helps