Facing an unexpected challenge.

We arrived a day before the final. When we arrived the hotel, we try to have a little rest but a phone call made it impossible. According to the state law of California, we can only do the demo inside some assigned park because there is strict law related to unmanned aircraft system. In this case, our demo location is the Sunnyvale Baylands Park. Unfortunately, there were not power line inside the park. We didn’t know what to do.

There was not time for us to be frustrated, we have to find some other way to demo our Automated Power Line Inspector app. We reached the park as soon as we can, and tried to find if there are some terrain similar to power lines or figure out some other solutions. We walked around the park and found a bridge, we decided to try our app with that bridge. We took off our Phantom3 and started the app. It was pretty successful but the bridge was too short that our demo would end in about 30 seconds and the lack of variation during the path could not demo our app to it’s extreme.

When we were testing with the bridge, some road’s line was detecting by our app. We thought it might be pretty impressive if our app can do road tracking as well. We tried to track the road and the result was less than satisfactory. We discussed together and decide to improve some part of algorithm in order to make the app adapt better when it comes to road tracking. Since the drone ran out of battery, we went back to hotel and modified algorithm while charging the battery.

Back to the park, the new algorithm worked more than perfect!!

Sunnyvale Baylands Park

Effort and passion always bring good news.

We are in the final of the competition! After a strict selection, DJI has chosen three teams around the country and we are so exciting to say we are one of them. The final will be hold in San Francisco, the final consist of two parts. First part is each team will give a 30 minutes presentations talking about the app or feature they have been working on and the challenges faced during the development. The second part is basically the demonstration of the app, to see if the function works properly. Our Automated Power Line Inspector already had a great response in analyzing images and track the power line, I believe the demo part would be a great chance for us! Please keep an eye on our blog, we will post about final's routine very soon. 

Testing IV

   

     Special thanks to our member Juan Wu who was the pilot of the shooting drone and also who edited the videos. In this video shows two cases which the first one was power line tracking and failure detection. The second shows the path selection when drone was in the intersection.

Testing III

     This time we wanted to try if our drone can flies properly calculating the mid-point of power lines and automatically follows the right direction. To demonstrate better the process, we used another drone to film the tested drone's route.

The golden one is for testing and the silver one is for filming

Here is the video on Youtube: 

     Because the battery of the drones only last approx. 20 mins so we couldn't finished the testing and the filming. We had a lots of problems before the filming of this video. First of all, we don't know what was happening with the golden drone which often had some trouble with it's camera that is always looking down before the take off and couldn't take off neither. This problem we had recently solved by updating the software. Secondly, as you can see in the video, the wind direction also affects the quality of shooting. Hoping a better outcome we are going to come back another day.

Process III

     The other day we failed our system in the detection of power lines on the road. So our team had made some improvements on the application mainly on the first image processing. The comparison below shows perfectly how our team could removed those noises made by the road: 

     Thanks to this development, at the term of detection of power lines could worked without any problem. As you guys can observe on the picture below, basically those power lines are well detected and some lines formed by some small noises are recognized too. 

     But those small lines are not effecting the final result of line detection. On this picture is shown the last development, the power lines are clearly noted and you can see a red circle and a blue line, each of them presents the midpoint of three power lines detected and the direction which the drone would followed. 

     To conclude, we are so exited about this brilliant result, we are not only done with the ground but also with the road now! That means that our application is almost done! Of course we are missing some testings more!

Guang(left) and Juan Wu(right) was preparing for the testing

Testing II

This is Guang testing with his iPhone and the remote control

     Previously we were testing our program on the soil, but this time we had chosen a place with the road shown. This summer in Texas is extremely hot, the sunlight burns and always when we go out for testing it's like a torture. However the drone can detect the power lines perfectly under this circumstances.

     The first picture shows the original image captured by the drone, with that on the program we can do the image processing to some important lines shown on the picture. After all, the program should recognize which of those are power liner. But this time, apparently the result was not ideal as we had expect. 

Summary of the system

This chart mostly shows our basic operation of the drone

Testing I

First outdoor testing on our program with the drone

   
     This time we are here to tell you our progress of the first program testing. First of all, there still existed some bugs on the application since every time when we turned the drone on. In general, when the camera starts to capture the image, it should send back at least 3 pictures per second to do the whole process of image, line extraction, etc.. But when we tested it can't even send back one image per minute. So after all we stopped the testing and thinking of where that we'd done was wrong.

Process II

    Can't wait to telling you guys, we had solve the noises caused by the ground on the back! As we can see on the image, comparing with the previous one this time we could successfully achieved our objective. Now, we are hoping our next meeting for the first testing on our program with the drone. :)

     This is the screenshot obtained after the evolution of our program

Process I

     We are excited to announce that we have handle remote control of the drone, the drone can use mathematical solving tools to do turn and figure graphical algorithms to track the power lines.

The original image(screenshot from a video)

Result from the detection of power lines

     We've also made a breakthrough in our project . We've successfully tested a mathematical calculation to identify some graphics of the cable, but those lines which are not interested for us also were detected as the second image have shown. We can observed how the noises produced by the grass had been effecting our final achievement. I hope we can resolve this problem as soon as possible.

Second Round

     We had successfully entered the second round of the DJI Developer Challenge! We also had received the Phantom 3 Professional from the DJI to continue working on our automatic power line inspector. We really appreciate this opportunity and we'll put the biggest effort to develop our aim of this project. In spite of the scarcity at beginning, we believed we can reach the best result about this research.

Our team working all together for the first time

Guang Zhou(Team Leader), Chao Wang and Celeste Wu

We are missing Wei Zhu and Juan Wu who are traveling at that time

DJI Developer Challenge

Join the Second DJI Developer Challenge and Win! | DJI

     Few months ago we joined the Second DJI Developer Challenge presenting our idea of totally Power Line CV inspector.
     The current power line is inspected by human, even the most advanced technology on the power line inspection, the aerial power line inspection service, also requires some experienced pilots  and electrical power line specialists to detect weak points before they impact the power line network. All of these methods are high cost and unsafe. So in this project we are trying to design a fully automated power line inspection methods by using current DJI Phantom 3 or Inspire 1(since the capability of real time HD video Stream).