Без темы <<  Mortality transition in Mexico, 1500-2000 Mrs.Zwahrs C.S.I. Class Landform Presentation  >> Input Video Step 1 - Reshape Image Step 1 - 8x8 Block Size Step 2 - Block Matrix Step 3 - PCA Projection Step 4 - Compute Score Step 5 - EV Matrix Detected Motion Detected Motion
Картинки из презентации «Motion Detection using PCA» к уроку английского языка на тему «Без темы»

Автор: Roland Miezianko. Чтобы познакомиться с картинкой полного размера, нажмите на её эскиз. Чтобы можно было использовать все картинки для урока английского языка, скачайте бесплатно презентацию «Motion Detection using PCA.ppt» со всеми картинками в zip-архиве размером 892 КБ.

## Motion Detection using PCA

содержание презентации «Motion Detection using PCA.ppt»
 Сл Текст Сл Текст 1 Motion Detection using PCA. Roland 16 normalized block matrix from Step 2 and Miezianko rmiezian@temple.edu Video project it onto the PCA projection matrix Analysis Project Spring 2004 Advisor: computed in Step 3 Only the first 3 PCA Prof. Dr. Longin Jan Latecki. 1. projections are used. 16. 2 Agenda. Motion Detection Input Video 17 Step 4 - Code. fileName = ...; Algorithm Steps (2-D and 3-D blocks) load(fileName, '-mat'); % matrix is BlockN Results Sample Videos and Results Video fileName = ...; load(fileName, '-mat'); % with 8x8 Detection Blocks Video with 32x32 matrix is pc Score = double(BlockN) * Detection Blocks Matlab Code Source Code. pc(:,1:3); 17. 2. 18 Step 5 - Compute EV. For each block 3 Input Video. MPEG video converted to sequence, load the PCA score matrix 2688 JPEG image frames Full RGB color. 3. computed in Step 4 Compute a covariance 4 Algorithm Steps. Reshape image to 8x8 matrix using a moving window of size 3 blocks Collect blocks from every frame, Compute eigenvalues (EVs) Sort to get the normalize and reshape array from 3-D 8x8 larges EV value Store the data in one EV blocks Compute PCA projection matrix per matrix, representing all blocks and all block Compute PCA score by projecting frames. 18. blocks from each frame onto that block’s 19 Step 5 - EV Matrix. EV matrix will 3-PCA projection Compute EV values with contain a single EV value for a W=3 for each block Generate global block-frame spatiotemporal location. 19. threshold based on all blocks and frames 20 Step 5 - Code. fileName = ...; Generate local dynamic threshold for each load(fileName, '-mat'); % matrix is Score block/frame with W=3 Generate motion dd = Score; evx = zeros(FrameEnd,3); W = matrix based on local and global dynamic 3; for i=W+1:length(dd)-W; cc = threshold for all blocks-frames. 4. dd(i-W:i+W,:); cm = cov(cc); 5 Step 1 - Details. Read the color image evx(i,:)=sort(eig(cm)'); end. 20. Resize the image by scale factor of 0.5 21 Step 6- Global Threshold. Load EV Convert the image to gray scale Reshape matrix from Step 5 Compute mean and the image into 8x8 distinct blocks standard deviation Find all entries in the Transpose and save the data Note: save per EV matrix that are below mean+2*std Update frame block data. 5. the EV matrix. 21. 6 Step 1 - Reshape Image. 6. 22 Step 6 - Code. fileName ...; 7 Step 1 - Code. fileName = ...; imN = load(fileName, '-mat'); % matrix is ev imread(fileName); imN = imresize(imN,0.5); gmean = mean(mean(ev')); gstd = imN = rgb2gray(imN); imCols = std(mean(ev')); LessThanThr = find(ev < im2col(imN,[bH bW],'distinct'); imT = (gmean+2*gstd)); ev(LessThanThr) = 20; 22. imCols'; 7. 23 Step 7 - Local Threshold. Use the 8 Step 1 - 8x8 Block Size. Block size updated EV matrix from Step 6 Compute a relative to image size Block 26x25 Image local dynamic threshold using window Size: 36x48 blocks. 8. Generate a Motion matrix of same size as 9 Step 2 - Collect Blocks. Collect same the EV matrix with a simple 0/1 values block from all the frames Create a single (1=motion). 23. matrix for each block location Reshape 24 Step 7 - Assumptions. Assume that vector from 3-D 8x8 blocks There are 1728 first 100 frames have no detectable motion matrices holding pixel values Each matrix Compute mean and std of first 100 frames is 2688 x 64 (frames x pixels/block). 9. for each block Compute local threshold for 10 Step 2-Normalize Blocks. Normalize each block using a moving window (W=3) each block by its mean value Each block Adjust local threshold, when no moving has its mean subtracted from each of the object is detected. 24. 64 pixel values Store the normalized block 25 Step 7 - Code 1. for BlockIndex = 1 : data to be used in Step 3 and Step 4. 10. NumBlocks W=3; current sample FrameStart = 11 Step 2 - Code. fileName = ...; 100; % first frames = no motion load(fileName, '-mat'); % matrix is BlockX meanl=mean(ev(BlockIndex,1:FrameStart)); % BlockN will be the normalized version of stdl=std(ev(BlockIndex,1:FrameStart)); BlockX BlockN = BlockX; BlockXMean = movingobject=0; for i = FrameStart+W : (mean(BlockX'))'; for FrameIndex = FrameEnd-W; ... (next slide) end % i end % FrameStart : FrameEnd BlockN(FrameIndex,:) BlockIndex. 25. = BlockX(FrameIndex,:) - 26 Step 7 - Code 2 (…). BlockXMean(FrameIndex,1); End % store lw=ev(BlockIndex,i-W:i); %left window normalized block matrix as BlockN. 11. rw=ev(BlockIndex,i+1:i+W); %right window 12 Step 2 - Block Matrix. Each block X of if mean(rw)-meanl>50*stdl 1728 total blocks has a matrix mobin(i)=mean(rw)-meanl; %moving obj representation of size 2688x64 Each block detected movingobject=1; is normalized by its mean value N = 2688. Motion(BlockIndex,i) = 1; else if 12. movingobject==0 13 Step 2 – 3-D 8x8 Blocks. Take 3 rows meanl=0.9*meanl+0.1*mean(lw); of Block matrix from previous slide 3x64 stdl=0.9*stdl+0.1*std(lw); end end. 26. Reshape into 1x192 vector 3-D blocks are 27 Step 8 - Motion Matrix. Motion matrix overlapping New 3-D Block Matrix is used is of size 1728x2688, same size as the EV in computing PCA scores and projection matrix It contains values 0 or 1, where 1 matrices. 13. = motion detected Use the Motion matrix to 14 Step 3 - Compute PCA. Load normalized create sample videos showing blocks where block matrix from Step 2 and compute the motion was detected. 27. PCA projection matrix for this block 28 Detected Motion. No motion Detected sequence Code: fileName = ...; Motion (red blocks). 28. load(fileName, '-mat'); % matrix is BlockN 29 Conclusion. The method of motion [pc,latent,explained] = detection using principal component pcacov(cov(BlockN)); 14. analysis combined with dynamic 15 Step 3 - PCA Projection. The principal thresholding yields very good results in components projection matrix contains 64 detecting motion Future projects will rows representing each pixel location in include processing images with variation the block and 64 columns representing 64 in size of the blocks. 29. principal components Only the first three 30 Questions & Answers. Sample Videos components are used in projection (first 3 8x8 Blocks 32x32 Blocks Principal columns). 15. Component Analysis Matlab Code. 30. 16 Step 4 - Compute Score. Load Motion Detection using PCA.ppt
http://900igr.net/kartinka/anglijskij-jazyk/motion-detection-using-pca-83841.html
cсылка на страницу

## Motion Detection using PCA

другие презентации на тему «Motion Detection using PCA»

«Teddy bear» - The modern genus Ursus appeared between five and ten million years ago. Some 38 million years ago bears began to go their own evolutionary way. Black bear. Teddy Bear. The size of the large males Kodiak Bear exceeds 3 m and their weight can reach a ton. Bears. Panda Bear. The grizzly bear male is on average 1.8 times as heavy as the grizzly bear female.

«At, in, on» - What are their names? -David and Paula. Напишите число на доске. It can be dangerous when children play football…the street. There is a dirty mark on the wall. I couldn’t see very well,because I was standing at the back. In the glass. On the ceiling on the wall on the floor on a page on her finger. Aims of the lesson.

«Жизнь в пустыне» - The Kalahari supports a variety of fauna and flora. Arabian desert. Cheetahs. The Gobi Desert is the coldest desert in the world. Hyenas. Turkestan Desert. The Kalahari Desert was once a much wetter place. People often travel by camel. The Kalahari Desert is a large arid sandy area in Southern Africa.

«April Fools Day» - In conclusion, I would say, when you think about how to make fun ofyou know, do not forget that the best joke-the one over which he laughs the loudest, played a joke on someone.  All Fool's Day. In Scotland the April fool's day lasts 48 hours.

«My family» - Iris is Richard`s mother. Айрис – мама (чья?) Ричарда. My mother is thirty three. Grandfather - дедушка. Grandmother - бабушка. Uncle - дядя. Father – папа. Possessive case – Притяжательный падеж имен существительных. Mother - мама. Sister - сестра. Meet my family. My brothers` names are Bill and Richard. Имена (чьи?) моих братьев – Бил и Ричард.

«Английский в начальной школе» - Нашим детям - единство радости, добра и красоты. Проектная деятельность. Проблема – поиск - решение. «Моделирование языковой среды при обучении английскому языку в начальной школе». Знание языка – обязательное требование жизни в современном обществе. Продуктивнее начинать обучение иностранному языку в начальной школе.

661 презентация
Урок

29 тем
Картинки