Hackaday Links: November 15, 2020

Now that we drive around cars that are more like mobile data centers than simple transportation, there’s a wealth of data to be harvested when the inevitable crashes occur. After a recent Tesla crash on a California highway, a security researcher got a hold of the car’s “black box” and extracted some terrifying insights into just how bad a car crash can be. The interesting bit is the view of the crash from the Tesla’s forward-facing cameras with object detection overlays. Putting aside the fact that the driver of this car was accelerating up to the moment it rear-ended the hapless Honda with a closing speed of 63 MPH (101 km/h), the update speeds on the bounding boxes and lane sensing are incredible. The author of the article uses this as an object lesson in why Level 2 self-driving is a bad idea, and while I agree with that premise, the fact that self-driving had been disabled 40 seconds before the driver plowed into the Honda seems to make that argument moot. Tech or not, someone this unskilled or impaired was going to have an accident eventually, and it was just bad luck for the other driver.

Last week I shared a link to Scan the World, an effort to 3D-scan and preserve culturally significant artifacts and create a virtual museum. Shortly after the article ran we got an email from Elisa at Scan the World announcing their “Unlocking Lockdown” competition, which encourages people to scan cultural artifacts and treasures directly from their home. You may not have a Ming Dynasty vase or a Grecian urn on display in your parlor, but you’ve probably got family heirlooms, knick-knacks, and other tchotchkes that should be preserved. Take a look around and scan something for posterity. And I want to thank Elisa for the link to the Pompeiian bread that I mentioned.

The Defense Advanced Research Projects Agency (DARPA)has been running an interesting challenge for the last couple of years: The Subterranean (SubT) Challenge. The goal is to discover new ways to operate autonomously below the surface of the Earth, whether for mining, search and rescue, or warfare applications. They’ve been running different circuits to simulate various underground environments, with the most recent circuit being a cave course back in October. On Tuesday November 17, DARPA will webcast the competition, which features 16 teams and their autonomous search for artifacts in a virtual cave. It could make for interesting viewing.

If underground adventures don’t do it for you, how about going upstairs? LeoLabs, a California-based company that specializes in providing information about satellites, has a fascinating visualization of the planet’s satellite constellation. It’s sort of Google Earth but with the details focused on low-earth orbit. You can fly around the planet and watch the satellites whiz by or even pick out the hundreds of spent upper-stage rockets still up there. You can lock onto a specific satellite, watch for near-misses, or even turn on a layer for space debris, which honestly just turns the display into a purple miasma of orbiting junk. The best bit, though, is the easily discerned samba-lines of newly launched Starlink satellites.

A doorbell used to be a pretty simple device, but like many things, they’ve taken on added complexity. And danger, it appears, as Amazon Ring doorbell users are reporting their new gadgets going up in flame upon installation. The problem stems from installers confusing the screws supplied with the unit. The longer wood screws are intended to mount the device to the wall, while a shorter security screw secures the battery cover. Mix the two up for whatever reason, and the sharp point of the mounting screw can find the LiPo battery within, with predictable results.

And finally, it may be the shittiest of shitty robots: a monstrous robotic wolf intended to scare away wild bears. It seems the Japanese town of Takikawa has been having a problem with bears lately, so they deployed a pair of these improbable looking creatures to protect themselves. It’s hard to say what’s the best feature: the flashing LED eyes, the strobe light tail, the fact that the whole thing floats in the air atop a pole. Whatever it is, it seems to work on bears, which is probably good enough. Take a look in the video below the break.

Continue reading “Hackaday Links: November 15, 2020”

Hacking A Cheap Action Cam Into A Dashcam With A Microcontroller

Repurposing commodity electronics is one of the true forms of hacking, and it’s always the simple little hacks that lead to big ones. [Everett] wanted to use a $20 GoPro clone as a dash cam, so he wired a microcontroller into it to automate some actions and make it practical.

The camera turns on automatically when connected to external power like a car charger, but starting and stopping a recording and power down all had to be done manually. [Everett] wanted to automate these functions, so he opened up the camera and started probing with an oscilloscope. He found the power button, record button, 3.3 V and external 5 V traces conveniently next to each other in the top of the camera.

To automate the required functions, he wired in a PIC10 on a small breakout board, powered by the 3.3 V line. It detects if 5 V is connected to the charging port on start-up via an N-channel FET, then automatically starts a recording. When the 5 V power is switched off with the car, it waits 10 seconds before stopping the recording and switching off the camera. If no external 5 V is not detected on start-up the microcontroller does nothing, which allows the camera to be used as a normal handheld. [Everett] mounted the camera to his rearview mirror with a magnetic bracket made using a combination of a 3D printer and 3D pen.

This is a simple and practical little hack, and the firmware is available on Github. Cheap dashcams are available for similar prices, but you won’t get any hacking satisfaction that way.

The very nature of actions cameras inspire hacking. You can simply add an external battery with the help of a 3D printer, or go all out and build a gimballed helmet cam from scratch

Reverse Engineering Liberates Dash Cam Video

If you’ve purchased a piece of consumer electronics in the last few years, there’s an excellent chance that you were forced to use some proprietary application (likely on a mobile device) to unlock its full functionality. It’s a depressing reality of modern technology, and unless you’re willing to roll your own hardware, it can be difficult to avoid. But [krishnan793] decided to take another route, and reverse engineered his DDPAI dash camera so he could get a live video stream from it without using the companion smartphone application.

Like many modern gadgets, the DDPAI camera creates its own WiFi access point that you need to connect to for configuration. By putting his computer’s wireless card into Monitor mode and running Wireshark, [krishnan793] was able to see that the smartphone was communicating with the camera using some type of REST API. After watching the clear-text exchanges for awhile, he not only discovered a few default usernames and passwords, but the commands necessary to configure the camera and start the video stream.

After hitting it with the proper REST messages, an nmap scan confirmed that several new services had started up on the device. Unfortunately, he didn’t get any video when he pointed VLC to the likely port numbers. At this point [krishnan793] checked the datasheet for the camera’s Hi3516E SoC and saw that it supported H.264 encoding. By manually specifying that as the video codec when invoking VLC, it was able to play a video stream from port 6200. A little later, he discovered that port 6100 was serving up the live audio.

Technically that’s all he wanted to do in the first place, as he was looking to feed the video into OpenCV for other projects. But while he was in the area, [krishnan793] also decided to find the download URL for the camera’s firmware, and ran it through binwalk to see what he could find out. Not surprisingly the security turned out to be fairly lax through the entire device, so he was able to glean some information that could be useful for future projects.

Of course, if you’d rather go with the first option and build your own custom dash camera so you don’t have to jump through so many hoops just to get a usable video stream, we’ve got some good news for you.

Homebrew Dash Cam Enables Full Suite Of Sensors

You heard it here first: dash cams are going to be the next must-have item for your daily driver. Already reaching market saturation in some parts of the world but still fairly uncommon in North America, we predict that car makers will soon latch onto the trend and start equipping cars with dash cams as standard equipment. And you can just bet that whatever watered-down, overpriced feature set they come up with will be sure to disappoint, so you might want to think about building your own Raspberry Pi dash cam with an accelerometer and lots of LEDS.

Still very much in the prototyping phase, [CFLanger]’s project is at its heart a dash cam, but it looks like he wants to go far beyond that. Raspivid and a PI NoIR camera take care of the video streaming, but the addition of a Pi SenseHAT gives [CFLanger] a bunch of options for sensing and recording the car’s environment. Not content with the SenseHAT’s onboard accelerometer, he added an ADXL345 to the sensor suite. The 64-pixel LED display is just for fun – it displays pitch and roll of the platform – and a yet-to-be-implemented bar-graph display will show acceleration in the X-axis. He figures the whole thing is good for a couple of days of video, but we hope he adds audio capture and perhaps ECU data from an OBDII-Bluetooth adapter.

We’ve seen surprisingly few DIY dash cams on Hackaday, at least so far. There has been a dash cam teardown and retasking, and there are plenty of dashboard computer builds, though. Seems like most hackers want that DIY self-driving car first.

Continue reading “Homebrew Dash Cam Enables Full Suite Of Sensors”