Vision Based Automated Cluster Tester
Vision Based Automated Instrument Cluster Tester: An automated machine vision system which acquires
the vehicle cluster (car dashboard) image using an USB interfaced webcam. The system further processes the
acquired image for predefined cluster configurations using computer vision algorithms developed and
implemented in C++. The extracted cluster data (readings) are displayed using VC++ (MFC) based GUI.
Comparing the given inputs and acquired image data, an instrument cluster can be tested from functional
point of view. [Used: VC++ (MFC), MATLAB]
Gesture Recognition using DSTW & DTW
Gesture Recognition using DSTW and DTW:
Scale and Translation invariant gesture recognition based on similarity measure between query
and model video sequences is implemented. Similarity measure is find out by using either simple
Dynamic Time Warping (DTW) or Advanced Dynamic Space Time Warping (DSTW) algorithm.
In presence of moving distractors motion and skin detection based hand detector module may
fail but use of dynamic space time warping algorithm (DSTW) to find the similarity measure
improves accuracy greatly over simple DTW algorithm.[Language Used: MATLAB]
Face Detector
Face image
Face image
Face detector:
Developed Face detector which uses Ada-Boosting and Bootstrapping algorithms at training
stage and multi-stage Face
of boosted classifiers aided with skin detection at testing stage.
[Language Used: MATLAB]
Face image
Person Tracker
Person partially in frame:20
Speed between frame:20 and frame:10 is 0 pixels per frame
Person present in POSE 2(Constricted Legs) inframe:77
Speed between frame:77 and frame:67 is 3.701351e+000 pixels per frame
Person present in POSE 1(Extended Legs) inframe:68
Speed between frame:68 and frame:58 is 3.733631e+000 pixels per frame
Person present in POSE 1(Extended Legs) inframe:85
Speed between frame:85 and frame:75 is 3.705401e+000 pixels per frame
Person Tracker: Implemented motion estimation based person tracker which draws bounding box around
person. The tracker determines center of mass of the person and calculates his/her speed from Euclidean
distance between centers of mass. Furthermore, the tracker determines walking pose of the person such as
Constricted Legs or Expanded Legs. [Language Used: MATLAB]
Face Recognition using Correlation,
Principal Component Analysis
and Ada-boosted Classifiers
100 eigenfaces
Rectangle Filter Classifier based adaptive boosting
Face Recognition using Correlation, Principal Component Analysis and Ada-boosted
A face recognition using simple 2-D cross correlation, PCA and Ada-Boosted Classifier(rectangle
filter) was implemented in order to compare accuracy and speed of these methods.[Language
Bayesian Probabilistic Skin Detector
Bayesian Probabilistic Skin Detector:
A simple Bayesian probabilistic skin detector using RGB histogram and Normalized RGB
histogram technique is implemented. The skin detector finds its application in face detection and
gesture recognition using skin detection based hand detector[Language Used: MATLAB]
Direction Identifier for maps
Direction Identifier:
A direction identifier based on canny edge detector is designed and implemented to determine
direction of prominent road in the given map image.[Language Used: MATLAB]
Digit Recognition
test20.bmp contains digit - 2
test15.bmp contains digit - 2
test10.bmp contains digit - 2
test40.bmp contains digit - 3
test35.bmp contains digit - 3
test31.bmp contains digit - 3
Digit Recognition:
A simple 2-D normalized cross correlation based digit recognition module is implemented. The
numerical digits are recognized even in presence of significant amount of noise.
[Language Used: MATLAB]

similar documents