Such a goal object, how to do contour detection and recognition, because shapes are various and dive-CodePudding

Like tobacco, a variety of shapes, I now is to put the overlap of separation, only independent, how to do this kind of target identification, contour extraction and calculate the size? Please help me about the reason thinking! I use opencv.

CodePudding user response:

You what things ah, binarization first

CodePudding user response:

reference 1st floor Xu Wenjun response:

you what things, first binarization

For instance, filtering, noise reduction, better adopt what way?

CodePudding user response:

the need to filter? Directly in cross section, and then find a suitable your histogram threshold, have the right job can cooperation,

CodePudding user response:

randomly choose the threshold value

CodePudding user response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and outline area calculation

CodePudding user response:

The

oh roar reference 5 floor? Response:

canny edge detection, noise reduction and binarization, then the contour detection, outline drawing and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

CodePudding user response:

refer to 6th floor ruan1978 response:

Quote: oh roar reference 5 floor? Response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

Feeling bad, after all, the noise reduction result is established on the basis of what as noise, can be all sorts of filtering are to try, and a nonlocal average filtering effect can also be, mainly the images as the input image filtering, can extract the outline, you want the degree have to adjust yourself

CodePudding user response:

The

oh roar reference 7 floor? Response:

Quote: refer to the sixth floor ruan1978 response:

Quote: oh roar reference 5 floor? Response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

Feeling bad, after all, the noise reduction result is established on the basis of what as noise, can be all sorts of filtering are to try, and a nonlocal average filtering effect can also be, mainly the images as the input image filtering, can extract the outline, you want the degree have to adjust yourself

He that I don't know he is sampling or on the production line test, but he can see the figure are mostly with a clap, guess he have a friend to do mechanical, he make software, visual personnel is missing but the middle part, because in a production environment, for the light source, camera, lens, Angle has certain requirements, the final image is not the same completely, so the images are far apart and eventually environment estimate, what's the point to the image filtering way?

CodePudding user response:

Can be used as a test, prototype builds with similar images, then the actual production line in debugging will be more convenient, if we can get the most close to the picture of the real production environment must be the best

CodePudding user response:

refer to the eighth floor Xu Wenjun response:

Quote: oh roar reference 7 floor? Response:

Quote: refer to the sixth floor ruan1978 response:

Quote: oh roar reference 5 floor? Response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

Feeling bad, after all, the noise reduction result is established on the basis of what as noise, can be all sorts of filtering are to try, and a nonlocal average filtering effect can also be, mainly the images as the input image filtering, can extract the outline, you want the degree have to adjust yourself

He that I don't know he is sampling or on the production line test, but he can see the figure are mostly with a clap, guess he have a friend to do mechanical, he make software, visual personnel is missing but the middle part, because in a production environment, for the light source, camera, lens, Angle has certain requirements, the final image is not the same completely, so the images are far apart and eventually environment estimate, what's the point to the image filtering way?

If it is the actual production line, the film should be fixed, and it is relatively stable, controlled by the mechanical can figure with a practical production line must be the best but if not, take other test images to study how to achieve the result that you want to build a more appropriate Demo, and then put in actual production line according to the actual situation debug

CodePudding user response:

The

references to the tenth floor roar? Response:

Quote: refer to the eighth floor Xu Wenjun response:

Quote: oh roar reference 7 floor? Response:

Quote: refer to the sixth floor ruan1978 response:

Quote: oh roar reference 5 floor? Response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

Feeling bad, after all, the noise reduction result is established on the basis of what as noise, can be all sorts of filtering are to try, and a nonlocal average filtering effect can also be, mainly the images as the input image filtering, can extract the outline, you want the degree have to adjust yourself

He that I don't know he is sampling or on the production line test, but he can see the figure are mostly with a clap, guess he have a friend to do mechanical, he make software, visual personnel is missing but the middle part, because in a production environment, for the light source, camera, lens, Angle has certain requirements, the final image is not the same completely, so the images are far apart and eventually environment estimate, what's the point to the image filtering way?

If it is the actual production line, the film should be fixed, and it is relatively stable, controlled by the mechanical can figure with a practical production line must be the best but if not, take other test pictures to research how to achieve the result that you want to build a more appropriate Demo, then on again according to the actual situation in practical production line debugging

Filtering is the purpose of eliminating the background value, but visual staff doing the experiment, the experimental background, lighting, Angle, camera, the lens will have certain requirements, even if the follow-up still need to filter, but imaging has been is not the same thing, of course, if only for this matter how do not understand, is really has a guiding significance,

CodePudding user response:

The

reference 11 floor Xu Wenjun reply:

Quote: reference to the tenth floor roar? Response:

Quote: refer to the eighth floor Xu Wenjun response:

Quote: oh roar reference 7 floor? Response:

Quote: refer to the sixth floor ruan1978 response:

Quote: oh roar reference 5 floor? Response:

Canny edge detection, noise reduction and binarization, then the contour detection, draw the outline and calculate the outline area

With what kind of noise reduction method is better? In view of the above graphics?

Feeling bad, after all, the noise reduction result is established on the basis of what as noise, can be all sorts of filtering are to try, and a nonlocal average filtering effect can also be, mainly the images as the input image filtering, can extract the outline, you want the degree have to adjust yourself

He that I don't know he is sampling or on the production line test, but he can see the figure are mostly with a clap, guess he have a friend to do mechanical, he make software, visual personnel is missing but the middle part, because in a production environment, for the light source, camera, lens, Angle has certain requirements, the final image is not the same completely, so the images are far apart and eventually environment estimate, what's the point to the image filtering way?

nullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnullnull