Autonomous Mobile Robot (Part 1): Overview & Hardware

Welcome to “Robot Boot Camp.” In this two-part article series, I’ll explain what you can do with a basic mobile machine, a few sensors, and behavioral programming techniques. Behavioral programming provides distinct advantages over other programming techniques. It is independent of any environmental model, and it is more robust in the face of sensor error, and the behaviors can be stacked and run concurrently.

My objectives for my recent robot design were fairly modest. I wanted to build a robot that could cruise on its own, avoid obstacles, escape from inadvertent collisions, and track a light source. I knew that if I could meet such objective other more complex behaviors would be possible (e.g., self-docking on low power). There certainly many commercial robots on the market that could have met my requirements. But I decided that my best bet would be to roll my own. I wanted to keep things simple, and I wanted to fully understand the sensors and controls for behavioral autonomous operation. The TOMBOT is the fruit of that labor (see Photo 1a). A colleague came up with the name TOMBOT in honor of its inventor, and the name kind of stuck.

Photo 1a—The complete TOMBOT design. b—The graphics display is nice feature.

In this series of articles, I’ll present lessons learned and describe the hardware/software design process. The series will detail TOMBOT-style robot hardware and assembly, as well as behavior programming techniques using C code. By the end of the series, I’ll have covered a complete behavior programming library and API, which will be available for experimentation.

DESIGN BASICS

The TOMBOT robot is certainly minimal, no frills: two continuous-rotation, variable-speed control servos; two IR (850 nm) analog distance measurement sensors (4- to 30-cm range); two CdS photoconductive cells with good lux response in visible spectrum; and, finally, a front bumper (switch-activated) for collision detection. The platform is simple: servos and sensors on the left and right side of two level platforms. The bottom platform houses bumper, batteries, and servos. The top platform houses sensors and microcontroller electronics. The back part of the bottom platform uses a central skid for balance between the two servos (see Photo 1).

Given my background as a Microchip Developer and Academic Partner, I used a Microchip Technology PIC32 microcontroller, a PICkit 3 programmer/debugger, and a free Microchip IDE and 32-bit complier for TOMBOT. (Refer to the TOMBOT components list at the end of this article.)

It was a real thrill to design and build a minimal capability robot that can—with stacking programming behaviors—emulate some “intelligence.” TOMBOT is still a work in progress, but I recently had the privilege of demoing it to a first grade class in El Segundo, CA, as part of a Science Technology Engineering and Mathematics (STEM) initiative. The results were very rewarding, but more on that later.

BEHAVIORAL PROGRAMMING

A control system for a completely autonomous mobile robot must perform many complex information-processing tasks in real time, even for simple applications. The traditional method to building control systems for such robots is to separate the problem into a series of sequential functional components. An alternative approach is to use behavioral programming. The technique was introduced by Rodney Brooks out of the MIT Robotics Lab, and it has been very successful in the implementation of a lot of commercial robots, such as the popular Roomba vacuuming. It was even adopted for space applications like NASA’s Mars Rover and military seekers.

Programming a robot according to behavior-based principles makes the program inherently parallel, enabling the robot to attend simultaneously to all hazards it may encounter as well as any serendipitous opportunities that may arise. Each behavior functions independently through sensor registration, perception, and action. In the end, all behavior requests are prioritized and arbitrated before action is taken. By stacking the appropriate behaviors, using arbitrated software techniques, the robot appears to show (broadly speaking) “increasing intelligence.” The TOMBOT modestly achieves this objective using selective compile configurations to emulate a series of robot behaviors (i.e., Cruise, Home, Escape, Avoid, and Low Power). Figure 1 is a simple model illustration of a behavior program.

Figure 1: Behavior program

Joseph Jones’s Robot Programming: A Practical Guide to Behavior-Based Robotics (TAB Electronics, 2003) is a great reference book that helped guide me in this effort. It turns out that Jones was part of the design team for the Roomba product.

Debugging a mobile platform that is executing a series of concurrent behaviors can be daunting task. So, to make things easier, I implemented a complete remote control using a wireless link between the robot and a PC. With this link, I can enable or disable autonomous behavior, retrieve the robot sensor status and mode of operations, and curtail and avoid potential robot hazard. In addition to this, I implemented some additional operator feedback using a small graphics display, LEDs, and a simple sound buzzer. Note the TOMBOT’s power-up display in Photo 1b. We take Robot Boot Camp very seriously.

Minimalist System

As you can see in the robot’s block diagram (see Figure 2), the TOMBOT is very much a minimalist system with just enough components to demonstrate autonomous behaviors: Cruise, Escape, Avoid, and Home. All these behaviors require the use of left and right servos for autonomous maneuverability.

Figure 2: The TOMBOT system

The Cruise behavior just keeps the robot in motion in lieu of any stimulus. The Escape behavior uses the bumper to sense a collision and then 180° spin with reverse. The Avoid behavior makes use of continuous forward-looking IR sensors to veer left or right upon approaching a close obstacle. The Home behavior utilizes the front optical photocells to provide robot self-guidance to a strong light highly directional source. It all should add up to some very distinct “intelligent” operation. Figure 3 depicts the basic sensor and electronic layout.

Figure 3: Basic sensor and electronic layout

TOMBOT Assembly

The TOMBOT uses the low-cost robot platform (ArBot Chassis) and wheel set (X-Wheel assembly) from Budget Robotics (see Figure 4).

Figure 4: The platform and wheel set

A picture is worth a thousand words. Photo 2 shows two views of the TOMBOT prototype.

Photo 2a: The TOMBOT’s Sharp IR sensors, photo assembly, and more. b: The battery pack, right servo, and more.

Photo 2a shows dual Sharp IR sensors. Just below them is the photocell assembly. It is a custom board with dual CdS GL5528 photoconductive cells and 2.2-kΩ current-limiting resistors. Below this is a bumper assembly consisting of two SPDT Snap-action switches with lever (All Electronics Corp. CAT# SMS-196, left and right) fixed to a custom pre-fab plastic front bumper. Also shown is the solderless breakout board and left servo. Photo 2b shows the rechargeable battery pack that resides on the lower base platform and associated power switch. The electronics stack is visible. Here the XBee/Buzzer and graphics card modules residing on the 32-bit Experimenter. The Experimenter is plugged into a custom carrier board that allows for an interconnection to the solderless breakout to the rest of the system. Finally, note that the right servo is highlighted. The total TOMBOT package is not ideal; but remember, I’m talking about a prototype, and this particular configuration has held up nicely in several field demos.

I used Parallax (Futaba) continuous-rotation servos. They use a three-wire connector (+5 V, GND, and Control).

Figure 5 depicts a second-generation bumper assembly.  The same snap-action switches with extended levers are bent and fashioned to interconnect a bumper assembly as shown.

Figure 5: Second-generation bumper assembly

TOMBOT Electronics

A 32-bit Micro Experimenter is used as the CPU. This board is based the high-end Microchip Technology PIC32MX695F512H 64-pin TQFP with 128-KB RAM, 512-KB flash memory, and an 80-MHz clock. I did not want to skimp on this component during the prototype phase. In addition the 32-bit Experimenter supports a 102 × 64 monographic card with green/red backlight controls and LEDs. Since a full graphics library was already bundled with this Experimenter graphics card, it also represented good risk reduction during prototyping phase. Details for both cards are available on the Kiba website.

The Experimenter supports six basic board-level connections to outside world using JP1, JP2, JP3, JP4, BOT, and TOP headers.  A custom carrier board interfaces to the Experimenter via these connections and provides power and signal connection to the sensors and servos. The custom carrier accepts battery voltage and regulates it to +5 VDC. This +5 V is then further regulated by the Experimenter to its native +3.3-VDC operation. The solderless breadboard supports a resistor network to sense a +9-V battery voltage for a +3.3-V PIC processor. The breadboard also contains an LM324 quad op-amp to provide a buffer between +3.3-V logic of the processor and the required +5-V operation of the servo. Figure 6 is a detailed schematic diagram of the electronics.

Figure 6: The design’s circuitry

A custom card for the XBee radio carrier and buzzer was built that plugs into the Experimenter’s TOP and BOT connections. Photo 3 shows the modules and the carrier board. The robot uses a rechargeable 1,600-mAH battery system (typical of mid-range wireless toys) that provides hours of uninterrupted operation.

Photo 3: The modules and the carrier board

PIC32 On-Chip Peripherals

The major PIC32 peripheral connection for the Experimenter to rest of the system is shown. The TOMBOT uses PWM for servo, UART for XBee, SPI and digital for LCD, analog input channels for all the sensors, and digital for the buzzer and bumper detect. The key peripheral connection for the Experimenter to rest of the system is shown in Figure 7.

Figure 7: Peripheral usage

The PIC32 pinouts and their associated Experimenter connections are detailed in Figure 8.

Figure 8: PIC32 peripheral pinouts and EXP32 connectors

The TOMBOT Motion Basics and the PIC32 Output Compare Peripheral

Let’s review the basics for TOMBOT motor control. The servos use the Parallax (Futaba) Continuous Rotation Servos. With two-wheel control, the robot motion is controlled as per Table 1.

Table 1: Robot motion

The servos are controlled by using a 20-ms (500-Hz) pulse PWM pattern where the PWM pulse can from 1.0 ms to 2.0 ms. The effects on the servos for the different PWM are shown in Figure 9.

Figure 9: Servo PWM control

The PIC32 microcontroller (used in the Experimenter) has five Output Compare modules (OCX, where X =1 , 2, 3, 4, 5). We use two of these peripherals, specifically OC3, OC4 to generate the PWM to control the servo speed and direction. The OCX module can use either 16 Timer2 (TMR2) or 16 Timer3 (TMR3) or combined as 32-bit Timer23 as a time base and for period (PR) setting for the output pulse waveform. In our case, we are using Timer23 as a PR set to 20 ms (500 Hz). The OCXRS and OCXR registers are loaded with a 16-bit value to control width of the pulse generated during the output period. This value is compared against the Timer during each period cycle. The OCX output starts high and then when a match occurs OCX logic will generate a low on output. This will be repeated on a cycle-by-cycle basis (see Figure 10).

Figure 10: PWM generation

Next Comes Software

We set the research goals and objectives for our autonomous robot. We covered the hardware associated with this robot and in the next installment we will describe the software and operation.

Tom Kibalo holds a BSEE from City College of New York and an MSEE from the University of Maryland. He as 39 years of engineering experience with a number of companies in the Washington, DC area. Tom is an adjunct EE facility member for local community college, and he is president of Kibacorp, a Microchip Design Partner.

Issue 265: Embedded Systems Abound

I recently read on CNN.com the transcript of an interview (May 9, 2002) with arachnologist Norman Platnick who stated: “You’re probably within seven or eight feet of spider no matter where you are. The only place on earth that has no spiders at all—as far as we know—is Antarctica.” It didn’t take long for me to start thinking about embedded systems and my proximity to them. Is the average person always within several feet of embedded systems? Probably not. But what about 50% or 60% of the time? E-mail me your thoughts.

Circuit Cellar 265, August 2012 - Embedded Development

Embedded systems are becoming ubiquitous. They’re in vehicles, mobile electronics, toys, industrial applications, home appliances, and more. If you’re indoors, the temperature is likely monitored and controlled by an embedded system. When you’re engaged in outdoor activities (e.g., hiking, golfing, biking, or boating), you probably have a few MCU-controlled devices nearby, such as cell phones, rangefinders, pedometers, and navigation systems. This month we present articles about how embedded systems work, and our authors also provide valuable insight about topics ranging from concurrency to project development.

Freescale’s Mark Pedley kicks off the issue with a revealing article about a tilt-compensating electronic compass (p. 16). Now you can add an e-compass to your next MCU-based project.

E-compass technology (Source: M. Pedley, CC265)

Turn to page 24 for an in-depth interview with Italy-based engineer Guido Ottaviani. His fascination with electronics engineering, and robotics in particular, will inspire you.

Have you ever come across a product that you know you could have designed better? Scott Weber had that experience and then acted on his impulse to build a more effective system. He created an MCU-based light controller (p. 32).

The MCU-based light controller is on the right (Source: S. Weber, CC265)

If you want to ensure a microcontroller works efficiently within one of your systems, you should get to know it inside and out. Shlomo Engelberg examines the internal structure of an I/O pin with a pull-up resistor (p. 40).

Bob Japenga continues his series “Concurrency in Embedded Systems” on page 44. He covers atomicity and time of check to time of use (TOCTTOU).

On page 48 George Novacek presents the second part of his series on project development. He covers project milestones and design reviews.

Ed Nisley’s June 2012 article introduced the topic of MOSFET channel resistance. On page 52 he covers his Arduino-based MOSFET tester circuitry and provides test results.

The MOSFET tester PCB hides the Arduino that runs the control program and communicates through the USB cable on the left edge. (Source: E. Nisley, CC265)

If you read Robert Lacoste’s June 2012 article, you now understand the basics of frequency mixers. This month he presents high-level design methods and tools (p. 58).

Jeff Bachiochi wraps up the issue with an examination of a popular topic—energy harvesting (p. 68). He covers PV cell technology, maximum power point tracking (MPPT), and charge management control.

A great way to investigate MPPT for your design is to use an STMicroelectronics evaluation board, such as this STEVAL-ISV006V2 shown in the top of the photo. The smaller cell in the center is rated at 165 mW (0.55-V output at 0.3 A) measuring 1.5” × 0.75”. At the bottom is a Parallax commercial-quality solar cell that is rated at 2.65 W (0.534-V output at 5.34 A) measuring 125 mm. (Source: J. Bachiochi, CC265)

Circuit Cellar 265 is currently on newsstands.

Propeller Games (P2): Game Logic

In the first part of this article series on Parallax Propeller-based gaming projects, I hooked up the hardware for the Hi/Lo game on a breadboard. Now I’ll write the game logic. The finished code is available here.

The power of the Propeller chip is in its multiple CPU cores. But you can do just fine with one processor, especially for a simple game like Hi/Lo. You program each of the processors in assembly or in the Parallax-invented SPIN high-level language. Assembly programs run blazingly fast directly in the CPU core. SPIN compiles to a binary format that is interpreted by the SPIN interpreter (written in assembly). The interpreter runs in the CPU core.

The CPU core is designed for speed, but it only has room for 512 instructions. The SPIN interpreter fetches your program byte by byte from shared RAM. Your code runs more slowly, but you have 32K of space to work with. I’ll use assembly in future projects, but SPIN is perfect for Hi/Lo.

A SPIN file is a collection of functions and data (shared by all functions in the file). The functions use local variables kept on a call stack. You break up your programming task into smaller functions that build on one another and call each other. You pass parameters to the functions and use the return values. It is all very similar to C programming though the syntax is different. The interpreter begins with the first function in your file no matter what you name it.

I started the project with a test “main” and the functions to control the Hi/Lo speaker, LEDs, and switches. 

This function plays a tone on the speaker (Source: C. Cantrell)

The “playTone” function generates a square wave on the speaker pin. The “cnt” register is a built-in 32-bit value that increments with every system clock. I run the prop stick full out with an 80-MHz clock configuration (5M-Hz crystal with a *16 internal multiplier). The “waitcnt” instruction puts the CPU to sleep until the system clock reaches the requested value. There are two waits in the loop that generates one clock cycle. Thus the generated frequency is roughly 40 MHz/freq. I say “roughly” because each instruction takes a little time to execute. The actual generated frequency is slightly less. There are much better ways to generate a precise square wave with the propeller hardware, but this is function is easy to understand, and it works fine for the simple Hi/Lo game.

The LED display is a collection of 14 segments and two dots that are turned on or off by writing a 1 or 0 to the Propeller port pins. The program use a look-up table that defines the various segment patterns to be shown.

The output pin bit patterns for numeric digits (Source: C. Cantrell)

The look-up table is defined in a data (DAT) section in the program. The SPIN language allows you to define binary constants with a “%” prefix. You can use the underscore (“_”) anywhere in any numeric constant to make it easier to read. The comment line just above the table shows how the segments map to bit positions in the propeller’s output register.

The “drawNumber” function displays a two digit value on the display. The function first divides the value by 10. The whole part (value/10) is the digit in the 10s place. The remainder (value//10) is the digit in the 1s place. The function looks up the separate patterns, ORs them together, and writes to the “outa” output register to toggle the lights.

I wrote LED functions to “drawBlank” (blank the display) and “drawHi” (show “Hi”) and “drawLo” (show “Lo”). These one-line functions are easy enough to code inline where they are used. But having the functions in distinct blocks makes the using code easier to understand and modify.

The functions to read the buttons return Boolean values: true if the switch is pressed or false if it is not. When a button is pressed, the corresponding input bit in “ina” goes to “1.” There are five buttons and five functions—one for each. There is also an “isAny” function to detect if any button is pressed.

The function returns "true" if a button is pressed. (Source: C. Cantrell)

The game itself has two distinct modes. The “splash” mode flashes “Hi/Lo” and waits for a player to press a button. This is an “attract” mode that draws players to the game. The “splash” function returns when a button has been pressed. The “playGame” function is the game logic. The function returns when the game is over. Thus the main loop simply calls the two functions in an infinite loop.

???????????. (Source: C. Cantrell)

The “splash” function calls “drawHi” and “drawLo” with a pause between.

The function attracts a player to the game. (Source: C. Cantrell)

The “pauseStopOnAnyButton” function counts up the delay and watches for “isAny”. It aborts the pause and returns true if a button is pressed. The “SPLASH_DELAY” is defined in the constant (“CON”) area of the program. I keep all “magic numbers” like delay counts and tone values in the CON area for easy tweaking.

The “playGame” function uses three helper functions: “getPlayerGuess,” “showWin,” and “showHint.” The “showWin” and “showHint” functions are just a couple of lines each and could be coded inline. Having them separate allows you to enhance the visual effects without changing the game logic code.

The “getPlayerGuess” does the real work of the game. It watches the buttons and changes the displayed number accordingly.

The function takes the player input. (Source: C. Cantrell)

The “getPlayerGuess” function is an infinite loop with five IF checks for each button. When the middle button is pressed the function returns with the global “playerGuess” variable holding the input value. The other buttons increment or decrement the digits on the display. Each IF block checks for overflow and plays a feedback tone.

There you have it: a simple Hi Lo game. The visual and input effects are in separate functions ready to be spruced up. I bet your solution has many more bells and whistles! I look forward to reading your ideas in the comments of this blog.

Next time I’ll wrap up the Hi Lo game with a little multitasking. I’ll write parallel programs to run in two new CPU cogs to manage sound effects and the LED display.

Chris Cantrell earned an MSEE from the University of Alabama. He writes Java and Flex for Emerson Network Power in Huntsville, Alabama. Circuit Cellar published 10 of his articles between 2002 and 2012: Issue 145, Issue 152, Issue 161, Issue 184, Issue 187, Issue 193, Issue 205, Issue 209, Issue 139, and Issue 260.

Propeller Games (P1): Hi Lo

Welcome to the Propeller Games! In a few installments, I’ll present several gaming projects that use the Parallax Propeller chip. The Propeller is perfect for gaming with its multiple CPU cores to handle simultaneous gaming activities and its on-board video generation circuitry.

My first game project is the classic “higher/lower” game, where the computer thinks of a number between 0 and 99 and you guess it. You have probably seen this played as the “Clock Game” on The Price is Right TV show, though some contestants struggle with a basic binary search algorithm. (You can watch videos of the game at YouTube.com.)

This entire project is built on a solderless breadboard. If you are new to the Propeller, this is the perfect project to get acquainted with the hardware and programming. If you are a Propeller guru, you will enjoy the nostalgia of gaming on LEDs and push buttons. Grab your breadboard and follow along.

Parts

What you’ll need:

  • Breadboard and wire
  • 9-VDC wall transformer
  • Parallax PropStick USB
  • Two-digit 7-segment LED display
  • Five SPST pushbuttons
  • Audio speaker
  • Sixteen 200-Ω resistors
  • Five 10-kΩ resistors

The board and basic parts

The Parallax Propeller chip requires a few external components. You need a 3.3-VDC power regulator, a crystal, and a USB-to-serial converter. You also need a serial EEPROM if you want the Propeller to run your program at power up. You can buy all these separately and wire them up on the breadboard. Or you can save time and space with the Parallax PropStick USB. It combines all these external parts on the same footprint as the 40-pin Propeller chip.

I bought the LED display for this project from Mouser Electronics (part number 630-HDSP-521E). The large red segments are common anode (common ground). You supply positive voltage from a propeller port pin through a 220-Ω resistor to light the segments.

I bought the push buttons from Pololu Robotics & Electronics (part number 1400). They are specially designed for mounting on a breadboard. One side of each switch is connected to 3.3 V and the other is connected to a propeller port pin and pulled to ground with a 10-kΩ resistor.

I bought the speaker from Digi-Key (part number 668-1140-ND). The negative terminal of the speaker hooks to the breadboard’s ground. The positive terminal hooks directly to a Propeller port pin.

A speaker, one LED segment, and one switch wired to the Propeller

I placed four of the switches on the corners of the display. These switches are used as up/down inputs for each digit allowing the player to select a number from 00 to 99. The fifth button to the right of the display is the “Enter” button.

The photo above shows the speaker, one LED segment, and one switch wired to the Propeller. I tested the hardware and software incrementally as I hooked it up instead of trying to debug the final system as a whole.

The Parallax Propeller Tool is the free graphical Integrated Development Environment (IDE) you use to develop code for the Propeller. The code editor colors and highlights your work making it easy to see functions and keywords. It also manages indentation. The SPIN programming language uses indentation to identify code blocks much as Python does.

Basic hardware test

The code above is my basic hardware test. The CON (constants) section at the top configures the clock speed of the chip: 5 MHz × 16 = 80 MHz. The OBJ (object) section pulls in the serial terminal driver library. This library object allows you to use the USB cable for both programming and an input/output terminal. The one second pause on line 12 gives you time to switch from the IDE program to the terminal program on your PC once the code is downloaded. The Propeller tool download includes the parallax serial terminal for your PC.

Line 10 sets general I/O pin 0 (P0) as an output (they are inputs by default). Line 17 reads the switch connected to P11 and turns the LED segment on or off accordingly. Line 18 prints the state of the input pins to the PC terminal in an infinite loop.

Parallax serial terminal

It took me a while to warm up to the SPIN programming language. It is syntactically very different from C and its derivatives. But conceptually it is familiar: you break your software up into functions and local/global variables. In the end the simplicity of the syntax and the friendliness of the IDE won me over!

I really like the “Propeller font” used in the Propeller Tool IDE. It includes special symbols you can use to draw circuits and timing diagrams in your code comments. For instance:

Check out the font

Now to wire up the rest of the LEDs and switches. I thought about wiring the left digit to the first port byte and the right digit to the second port byte so that the segments are laid out the same way in each byte. This would make the software easier to write. But the pins for the segments on the display are kind of scattered around at random. The wiring is easier and neater if you wire the segments from the bottom of the display to the bottom of the propeller and from the top of the display to the top of the propeller. You can make up for the scattered pattern with software.

Two tips: Wire the segments from the bottom of the display to the bottom of the Propeller. Wire from the top of the display to the top of the Propeller.

Hi/Lo breadboard layout

That’s it for this installment. Now I’ll clean up all the little wire stripping sprinkles I left around my workbench. In Part 2 of this series, I’ll switch modes from hardware to software and write the Hi/Lo game. Hopefully you are following along. Until next time, may the COGs be ever in your favor.

Chris Cantrell earned an MSEE from the University of Alabama. He writes Java and Flex for Emerson Network Power in Huntsville, Alabama. Circuit Cellar published 10 of his articles between 2002 and 2012: Issue 145, Issue 152, Issue 161, Issue 184, Issue 187, Issue 193, Issue 205, Issue 209, Issue 139, and Issue 260.

Robot Design with Microsoft Kinect, RDS 4, & Parallax’s Eddie

Microsoft announced on March 8 the availability of Robotics Developer Studio 4 (RDS 4) software for robotics applications. RDS 4 was designed to work with the Kinect for Windows SDK. To demonstrate the capabilities of RDS 4, the Microsoft robotics team built the Follow Me Robot with a Parallax Eddie robot, laptop running Windows 7, and the Kinect.

In the following short video, Microsoft software developer Harsha Kikkeri demonstrates Follow Me Robot.

Circuit Cellar readers are already experimenting Kinect and developing embedded system to work with it n interesting ways. In an upcoming article about a Kinect-based project, designer Miguel Sanchez describes a interesting Kinect-based 3-D imaging system.

Sanchez writes:

My project started as a simple enterprise that later became a bit more challenging. The idea of capturing the silhouette of an individual standing in front of the Kinect was based on isolating those points that are between two distance thresholds from the camera. As depth image already provides the distance measurement, all the pixels of the subject will be between a range of distances, while other objects in the scene will be outside of this small range. But I wanted to have just the contour line of a person and not all the pixels that belong to that person’s body. OpenCV is a powerful computer vision library. I used it for my project because of function blobs. This function extracts the contour of the different isolated objects of a scene. As my image would only contain one object—the person standing in front of the camera—function blobs would return the exact list of coordinates of the contour of the person, which was what I needed. Please note that this function is a heavy image processing made easy for the user. It provides not just one, but a list of all the different objects that have been detected in the image. It can also specify is holes inside a blob are permitted. It can also specify the minimum and maximum areas of detected blobs. But for my project, I am only interested in detecting the biggest blob returned, which will be the one with index zero, as they are stored in decreasing order of blob area in the array returned by the blobs function.

Though it is not a fault of blobs function, I quickly realized that I was getting more detail than I needed and that there was a bit of noise in the edges of the contour. Filtering out on a bit map can be easily accomplished with a blur function, but smoothing out a contour did not sound so obvious to me.

A contour line can be simplified by removing certain points. A clever algorithm can do this by removing those points that are close enough to the overall contour line. One of these algorithms is the Douglas-Peucker recursive contour simplification algorithm. The algorithm starts with the two endpoints and it accepts one point in between whose orthogonal distance from the line connecting the two first points is larger than a given threshold. Only the point with the largest distance is selected (or none if the threshold is not met). The process is repeated recursively, as new points are added, to create the list of accepted points (those that are contributing the most to the general contour given a user-provided threshold). The larger the threshold, the rougher the resulting contour will be.

By simplifying a contour, now human silhouettes look better and noise is gone, but they look a bit synthetic. The last step I did was to perform a cubic-spline interpolation so contour becomes a set of curves between the different original points of the simplified contour. It seems a bit twisted to simplify first to later add back more points because of the spline interpolation, but this way it creates a more visually pleasant and curvy result, which was my goal.

 

(Source: Miguel Sanchez)
(Source: Miguel Sanchez)

The nearby images show aspects of the process Sanchez describes in his article, where an offset between the human figure and the drawn silhouette is apparent.

The entire article is slated to appear in the June or July edition of Circuit Cellar.