Help with issues regarding Hough line detection and spaced characters.

Hi, I need to develop an application that takes image of digital circuit as input and prepares RTL code as output. For that I need to separate i/p & o/p labels, logic gates, wirelines connecting gates in order to analyze connections and prepare RTL code accordingly.

1. First labels are recognized & eliminated from binary type image using ocr() as stated in answer to Question 1 . As the labels can contain multiple characters such as A1, ZD, Cin, Cout etc.. , how to check easily if any two labels are located closely for example label 'ZD' in fig shown in Question 2 .

2. Recognize all straight lines (or logic gates, anything first) that represent wire connection between gates, and gates & labels. The main issue here is use of dots and line crossings. The line crossings are to be represented in a small semicircle to denote no connection between lines, while dot represent a connection. When Hough transform is used, it takes care of this criteria by the way it detects wirelines with dots and line crossings (see the figure below that shows Hough recognized wirelines' starting point in yellow, end point in red, and the wirelines contain both dots and line crossings).

I am looking exactly for such solution, but it doesn't detect all the wirelines. The Hough tranform code for detecting lines is posted in this question .So, how can I make Hough transform detect all the wirelines or is there any solution that detects the wirelines in the same way, and is more efficient than Hough transform to detect all the wirelines.

I used the same code exactly as provided in this example to mark start and end points in yellow & red. I really couldn't understand why the lines marked in green blocks are not detected. Someone please suggest how I can achieve 100% recognition.