Object Detection

@davee I think there is merit in pursuing your line of thinking. As I have pointed out using a conventional lens causes several problems and the 'line scanner' eliminates them. It remains to be seen if that device is sufficiently fast enough and sensitive enough to detect the defect going by at 1 to 2 metres a minute. In the example he showed earlier, it looks like a backlighting scenario may be what @wisam is thinking. I can't tell since we still do not have an understanding of what a defect is.

It seems to me this is an industrial project and there is very little we can add to it but maybe I am mistaken.

@davee @zander

It is unclear to me what @wisam expects from this forum?

The sparsity of detail on what he has or is wanting to do doesn't help and everyone is giving "answers" based on guesses of what it is all about. Spoon feeding information over a long exchange of posts, something I have seen happen before, I just find annoying.

Ron wrote: "... it looks like a backlighting scenario may be what @wisam is thinking." is an example of this guess what is going on here as a result of a lack of detail.

There appears to be solutions else where so I would suggest @wisam do a google search.

https://www.isravision.com/en-en/industries/plastic-film-foil-sheets

and so on ...

@robotbuilder Backlighting wasn't a guess, he showed it in action. However, I agree this forum is not what he needs, this is an industrial situation, not a hobbyist. He needs to look for a company to solve his problem.

@zander @davee @ron @robotbuilder 

We are a CCTV and networks company, and this matter was presented to us by the owners of one of the factories. We contacted a company specialized in industrial visions, after a week-long study, they suggested developing software to detect defects and send alert, but the surprise was at the cost of the software - $60,000, so we resorted to alternative solutions.

@wisam Ok, thanks you for your honesty. 

We are not a free or a fee based consultancy, we are hobbyists.

$60,000 is cheap, something like 1 person month or less. Take the deal!!!

Based on at least a 5 yr payback, that's $1,000 a month or roughly $250 a week, $50 a day. or $6.25 an hour. That is about 1/3 minimum wage here in Canada.

Just an FYI, when I last worked as a consultant, I charged close to $200 USD an hour, with inflation that would now be at least $330. That is $2,640 a day, $13,200 a week, $52,800 for 4 weeks.

The $60,000 is a bargain although it will probably go up even as much as double or more if you do not communicate well with the consultants, and based on what has transpired here, that is guaranteed to happen. But amortize it over 5 years and you will see how cheap it is.

@wisam 

It seems to me the factory itself needs to contact people that specialize in industrial vision systems. They would have the hardware and programmers capable of writing the code to run on their systems. An algorithm would have to be written or neural net trained on thousands of examples to recognize the defects and then integrated into the software environment of their vision systems. Maybe it costs $60,000 to employ a programmer with the knowledge and coding skills to realize a solution.

I doubt your cctv cameras would be up to the task of capturing images at high speed? These images would have to be sent to the computer that does the image processing. I for example use webcams to capture images to process them using a pc but that would not be fast enough for the application you have presented.

@robotbuilder Just one point of clarification, a programmer qualified for this job would be earning in the low to mid 6 digits. College kids working summer jobs make $40,000 annualized, it's 2024, not 1980.

Also, the actual coding part is about 20% of the project cost at most.

@zander 

In the original post by @wisam he did state that they were a surveillance company and a client wondered if they could use their system to monitor the plastic for defects at the same time!

"when detected by the surveillance camera that will be installed simultaneously with the applied solution."

My fault for not reading the post properly for which I can only apologize. A simple "not practical/possible would probably have been the best answer.

@robotbuilder ???? The camera he is talking about is a license plate reader, not a simple surveillance camera everyone has at home etc.

I just realized this is even worse than I thought.

I thought he represented the plastic film company and was looking to install a defect detection system. It turns out that is somebody else who has asked him if his CCTV system can be turned into a defect detection system. He then contacted a vision company but he is basking at $60,000 to pay for some development. That is dirt cheap, only 1 person month

He is trying to use us to help him invent a product that he will get paid handsomely for. 

Can you imagine in your former job asking the internet to do your job for you? If I had done that they would fire me so fast.

@robotbuilder Yes, his poor communication skills have wasted a lot of our time. Time to say good riddance to ...... well you know the rest. Now where is the unsubscribe button?

@wisam Hi, I've been following this post thread with some interest as I have seen this type of system employed in industrial manufacturing, specifically paper mills where I worked for 30 years. The concept of camera line scanners is what is mostly used, and I have included a link to an article I found that may help you better understand how this is done..

https://www.imveurope.com/white-paper/understanding-line-scan-camera-applications

Hopefully, you will get an idea of the pros and cons of using the cameras for high speed defect detection. will I'm retired now, the big machine at our mill was running @ 3450 fpm...or 57.5 fps,  that's equivalent to 17.53 mps Metes per second !  Good luck with your project.. and keeps us informed of your progress..

kind regards,

LouisR

@inst-tech FYI @wisam PERFECT. Everything I was concerned about is answered in that product and the approach is not something I thought of or in my realm of experience but I can see immediately it is the right engineering solution. This is the forum at it's best!

I expect the cost to the end user will be many times the $60,000 that @wisam thought was too much. Sorry, you just don't understand cost/benefit accounting nor industrial costs. Spending possibly several million for this class of scanner is a no-brainer if it reduces waste by only 10% of annual sales measured in the 10's of millions. A payback of 10 years is quite acceptable in my experience. The likelihood is it will reduce waste by the high 90's so payback will be almost instantaneous.

@wisam you now have at least a moral if not legal responsibility to report this back to the end user and this topic needs to be marked 'solved'.

@wisam I sent you a Private Message