将VS2017+OpenCV3.4.1下上个文章说写的不完整Retinex代码处理结果图拿到matlab,用matlab中下载的yolo算法进行目标检测。
matlab版本为2020a
在送入yolo网络之前,加一个中值滤波,再进行目标检测。
%% Object Detection Using YOLO v2 Object Detector
% The following code demonstrates running prediction on a pre-trained YOLO v2
% network, trained on COCO dataset.
%% *Download the pre-trained network*
modelName = 'tinyYOLOv2-coco';
helper.downloadPretrainedYOLOv2(modelName);
pretrained = load(modelName);
detector = pretrained.yolov2Detector;
% Detect Objects using YOLO v2 Object Detector
% Read test image.
%img = imread('sherlock.jpg');
img = imread('FLIR_03942.jpg');
% Detect objects in test image.
[boxes, scores, labels] = detect(detector, img);
% Visualize detection results.
img = insertObjectAnnotation(img,'rectangle',boxes,labels);
figure,
% subplot(1,2,1),
imshow(img);
%%
%-------------------中值滤波+ yolo----------------------------------
%
modelName = 'tinyYOLOv2-coco';
helper.downloadPretrainedYOLOv2(modelName);
pretrained = load(modelName);
detector = pretrained.yolov2Detector;
% Detect Objects using YOLO v2 Object Detector
% Read test image.
img = imread('03942_r.jpg');
n=3; %模板大小
[height, width]=size(img);
x1=double(img); %数据类型转换
x2=x1; %转换后的数据赋给x2
for i=1:height-n+1
for j=1:width-n+1
c=x1(i:i+(n-1),j:j+(n-1)); %在x1中从头取模板大小的块赋给c
e=c(1,:); %e中存放是c矩阵的第一行
for u=2:n %将c中的其他行元素取出来接在e后使e为一个行矩阵
e=[e,c(u,:)];
end
med=median(e); %取一行的中值
x2(i+(n-1)/2,j+(n-1)/2)=med; %将模板各元素的中值赋给模板中心位置的元素
end
end
img=uint8(x2); %未被赋值的元素取原值
%%
% Detect objects in test image.
[boxes, scores, labels] = detect(detector, img);
% Visualize detection results.
img = insertObjectAnnotation(img,'rectangle',boxes,labels);
% subplot(1,2,2),
figure,imshow(img);
处理后的结果
对比没有retinex+中值滤波的
以上仅发布了部分代码