【图像分割】光流生成标签（matlab）

本文主要是介绍【图像分割】光流生成标签（matlab），希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

1. 框架

2. opticalFlow_label

close all; clear; clc;
% 使用光流进行标签的生成
%% 视频帧的读取
npy_data = readNPY('train.npy');%% 提取标签的坐标
first_label = squeeze(npy_data(2,1,:,:));
h = fspecial("gaussian", [3, 3], 1);
first_label_g = imfilter(first_label, h, 'replicate');   % 'conv'
first_label_c = edge(first_label_g, "canny");
% imshow(uint8(first_label_c*255));
first_label_double = im2double(first_label_c);
first_label_bw = im2bw(first_label_double, 0.5);
% imshow(uint8(first_label_bw * 255));
[h, w] = size(first_label);
xPos = [];
yPos = [];
step = 0;
for i = 1:hfor j = 1:wif first_label_bw(i, j) == 1
%             xPos = [xPos, i];       % 保存为行
%             yPos = [yPos, j];step = step + 1;if mod(step, 1) == 0xPos = [xPos; j];   % 保存为列yPos = [yPos; i];endendend
end%% 逐帧处理
first_frame = squeeze(npy_data(1,1,:,:));
first_frame = uint8(first_frame);
% imshow(first_frame);
[c,frame_num,img_h,img_w] = size(npy_data);
num = 0;
save_npy(1,1,:,:) = first_frame;
save_npy(2,1,:,:) = first_frame;   % 预留一个通道，用于保存标签for i = 2  % 2:frame_numnum = num + 1;currFrame = squeeze(npy_data(1,i,:,:));currFrame = uint8(currFrame);pyramidLayer = 4;kernelSize = 3;sigma = 1.5;iterNumMax = 5;ww = 13;[u, v] = affineLKopticalFlow(first_frame, currFrame, xPos, yPos, pyramidLayer, kernelSize, sigma, iterNumMax, ww);% 显示newFrame = repmat(currFrame, [1, 1, 3]);new_xPos = xPos + u;new_yPos = yPos + v;newFrame1 = zeros(size(currFrame));for kk = 1:length(xPos)%显示newFrame(int16(new_yPos(kk)), int16(new_xPos(kk)), 1) = 255;newFrame(:, :, 2) = currFrame;newFrame(int16(new_yPos(kk)), int16(new_xPos(kk)), 3) = 0;newFrame1(int16(new_yPos(kk)), int16(new_xPos(kk))) = 1;end   save_npy(1,i,:,:) = currFrame;save_npy(2,i,:,:) = first_frame;
%     writeNPY(save_npy, "test.npy");figure(1)imshow(newFrame)title("Pre")% 形态学处理 离散点的连接se = strel('disk',51);   % 加大半径 可以进行填充fc = imclose(newFrame1, se);
%     imwrite(fc, "fc.jpg")% bw = im2bw(fc);% fill_img = imfill(bw, "holes");   % 封闭区域figure(2)imshow(fc)title("fc")%将mask显示在图片上mask_frame = repmat(currFrame, [1, 1, 3]);[mask_h, mask_w, channel] = size(mask_frame);for i = 1:mask_hfor j = 1:mask_wif fc(i, j) == 1mask_frame(i, j, 1) = 255;mask_frame

这篇关于【图像分割】光流生成标签（matlab）的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！

---