OpenCV 3 Blueprints Table of Contents OpenCV 3 Blueprints Credits About the Authors About the Reviewers www.PacktPub.com Support files, eBooks, discount offers, and more Why subscribe? Free access for Packt account holders Preface What this book covers What you need for this book Basic installation guides Who this book is for Conventions Reader feedback Customer support Downloading the example code Downloading the color images of this book Errata Piracy Questions 1. Getting the Most out of Your Camera System Coloring the light Capturing the subject in the moment Rounding up the unusual suspects Supercharging the PlayStation Eye Supercharging the ASUS Xtion PRO Live and other OpenNI-compliant depth cameras Supercharging the GS3-U3-23S6M-C and other Point Grey Research cameras Shopping for glass Summary 2. Photographing Nature and Wildlife with an Automated Camera Planning the camera trap Controlling a photo camera with gPhoto2 Writing a shell script to unmount camera drives Setting up and testing gPhoto2 Writing a shell script for exposure bracketing Writing a Python script to wrap gPhoto2 Finding libgphoto2 and wrappers Detecting the presence of a photogenic subject Detecting a moving subject Detecting a colorful subject Detecting the face of a mammal Processing images to show subtle colors and motion Creating HDR images Creating time-lapse videos Further study Summary 3. Recognizing Facial Expressions with Machine Learning Introducing facial expression recognition Facial expression dataset Finding the face region in the image Extracting the face region using a face detection algorithm Extracting facial landmarks from the face region Introducing the flandmark library Downloading and compiling the flandmark library Detecting facial landmarks with flandmark Visualizing the landmarks in an image Extracting the face region Software usage guide Feature extraction Extracting image features from facial component regions Contributed features Advanced features Visualizing key points for each feature type Computing the distribution of feature representation over k clusters Clustering image features space into k clusters Computing a final feature for each image Dimensionality reduction Software usage guide Classification Classification process Splitting the dataset into a training set and testing set Support vector machines Training stage Testing stage Multi-layer perceptron Training stage Define the network Train the network Testing stage K-Nearest Neighbors (KNN) Training stage The testing stage Normal Bayes classifier Training stage Testing stage Software usage guide Evaluation Evaluation with different learning algorithms Evaluation with different features Evaluation with a different number of clusters System overview Further reading Compiling the opencv_contrib module Kaggle facial expression dataset Facial landmarks What are facial landmarks? How do you detect facial landmarks? How do you use facial landmarks? Improving feature extraction K-fold cross validation Summary 4. Panoramic Image Stitching Application Using Android Studio and NDK Introducing the concept of panorama The Android section – an application user interface The setup activity layout Capturing the camera frame Using the Camera API to get the camera frame Implementing the Capture button Implementing the Save button Integrating OpenCV into the Android Studio Compiling OpenCV Android SDK to the Android Studio project Setting up the Android Studio to work with OpenCV Importing the OpenCV Android SDK Creating a Java and C++ interaction with Java Native Interface (JNI) Compiling OpenCV C++ with NDK/JNI Implementing the OpenCV Java code Implementing the OpenCV C++ code Application showcase Further improvement Summary 5. Generic Object Detection for Industrial Applications Difference between recognition, detection, and categorization Smartly selecting and preparing application specific training data The amount of training data Creating object annotation files for the positive samples Parsing your positive dataset into the OpenCV data vector Parameter selection when training an object model Training parameters involved in training an object model The cascade classification process in detail Step 1 – grabbing positive and negative samples Step 2 – precalculation of integral image and all possible features from the training data Step 3 – firing up the boosting process Step 4 – saving the temporary result to a stage file The resulting object model explained in detail HAAR-like wavelet feature models Local binary pattern models Visualization tool for object models Using cross-validation to achieve the best model possible Using scene specific knowledge and constraints to optimize the detection result Using the parameters of the detection command to influence your detection result Increasing object instance detection and reducing false positive detections Obtaining rotation invariance object detection 2D scale space relation Performance evaluation and GPU optimizations Object detection performance testing Optimizations using GPU code Practical applications Summary 6. Efficient Person Identification Using Biometric Properties Biometrics, a general approach Step 1 – getting a good training dataset and applying application-specific normalization Step 2 – creating a descriptor of the recorded biometric Step 3 – using machine learning to match the retrieved feature vector Step 4 – think about your authentication process Face detection and recognition Face detection using the Viola and Jones boosted cascade classifier algorithm Data normalization on the detected face regions Various face recognition approaches and their corresponding feature space Eigenface decomposition through PCA Linear discriminant analysis using the Fisher criterion Local binary pattern histograms The problems with facial recognition in its current OpenCV 3 based implementation Fingerprint identification, how is it done? Implementing the approach in OpenCV 3 Iris identification, how is it done? Implementing the approach in OpenCV 3 Combining the techniques to create an efficient people-registration system Summary 7. Gyroscopic Video Stabilization Stabilization with images Stabilization with hardware A hybrid of hardware and software The math The camera model The Camera motion Rolling shutter compensation Image warping Project overview Capturing data Recording video Recording gyro signals Android specifics Threaded overlay Reading media files Calibration Data structures Reading the gyroscope trace The training video Handling rotations Rotating an image Accumulated rotations The calibration class Undistorting images Testing calibration results Rolling shutter compensation Calibrating the rolling shutter Warping with grid points Unwarping with calibration What’s next? Identifying gyroscope axes Estimating the rolling shutter direction Smoother timelapses Repository of calibration parameters Incorporating translations Additional tips Use the Python pickle module Write out single images Testing without the delta Summary Index