0:00:16 | thank you for introduction basic uh my name is wendy no postdoc from ghent university |
---|
0:00:24 | that they i'm gonna put in that use something about this classification of hyperspectral image |
---|
0:00:31 | oh would happen easily based on extended morphological profiles responsive reconstructions |
---|
0:00:39 | different from the at normal actually image hyperspectral sensor cup two dates bits hundred append |
---|
0:00:47 | but it's pixels |
---|
0:00:50 | hot very hot it's hot continuing with reflections different materials have different spectral reflections |
---|
0:01:00 | and we can identify different materials according to its reflections for instance here |
---|
0:01:10 | according to this spectrum reflects the reflections we notice these soil and the this reflections |
---|
0:01:18 | basis what the |
---|
0:01:21 | for some hyperspectral image |
---|
0:01:25 | the spectral reflections of some of pets are similar when only using spectral information to |
---|
0:01:35 | do classification we cannot get sucked is final result spatial information it's very important spatial |
---|
0:01:43 | information such as object size shape and texture it's very important to improve the classification |
---|
0:01:51 | result |
---|
0:01:52 | when they called by |
---|
0:01:55 | both spectral information and the spatial information we can improve the classification accuracy a lot |
---|
0:02:05 | this is all light of my presentation today first i will give a lot of |
---|
0:02:11 | you about feature extractions at the morphological profiles which are widely used to explore ohms |
---|
0:02:20 | the spectral and spatial information of hyperspectral image then i mean in that use all |
---|
0:02:26 | propose it can be morphological profiles we sparse reconstructions and show some experimental results of |
---|
0:02:35 | the last but not least it conclusion and future work |
---|
0:02:42 | hyperspectral image it's difficult to process due to its high dimensionality |
---|
0:02:49 | feature extraction |
---|
0:02:52 | it's use it's always used as preprocessing to reduce the dimensionality and the lead then |
---|
0:02:59 | there's the of the hyperspectral states |
---|
0:03:04 | uh the |
---|
0:03:06 | pretty simple components and the cannot pretty simple can note component devices are most widely |
---|
0:03:12 | my set used to do basic directions |
---|
0:03:18 | principal component analysis extracted features on which the tomb melody alliance |
---|
0:03:27 | it's not seem to maximize |
---|
0:03:31 | the first few extracted features can you percent most of a information of heidi by |
---|
0:03:38 | original hyperspectral states and the last are mainly noise for instance here |
---|
0:03:46 | the first we extracted pretty simple components can be presented ninety nine percent of Q |
---|
0:03:53 | melody but lines of this one hundred and three bands hyperspectral target |
---|
0:04:01 | and the so as the identity i dependence component and |
---|
0:04:08 | to achieve the same ninety nine percent |
---|
0:04:12 | Q melody lines can opening simple component needs poles |
---|
0:04:19 | twelve can uh features extracted features but this as a nonlinear mess that's kernel principal |
---|
0:04:26 | component the licensees can proceed that non linear distribution of that eight |
---|
0:04:33 | and uh you see even on the total extracted features it's have some detail features |
---|
0:04:40 | but for pretty simple and the linear ica in the in its twelve extracted features |
---|
0:04:48 | maybe the noise |
---|
0:04:54 | uh spatial information it's important especially for the high low solution up and hyperspectral image |
---|
0:05:01 | in particular morphological profiles it's one of the most popular myself which are used to |
---|
0:05:07 | explore the spatial information of hyperspectral image |
---|
0:05:12 | that's take morphological closing as example to see how morphological profiles pretty C the spatial |
---|
0:05:20 | information of hyperspectral date |
---|
0:05:24 | in morphological closings a structure it demands are moved in inside the image when there's |
---|
0:05:33 | the size of the dark object it's more that then these structural elements |
---|
0:05:42 | then the object will be removed |
---|
0:05:46 | at the size of this structure demands increase more and more dark |
---|
0:05:54 | small object uh disappear in the disappear in this image |
---|
0:06:01 | and then the size of a small object only like it this the size of |
---|
0:06:07 | the strict demons |
---|
0:06:11 | so we say that's morphological profiles contain yeah yes thank you |
---|
0:06:19 | uh yeah thank you thank you much yeah |
---|
0:06:23 | and the this size it's like you like it to the size of the structure |
---|
0:06:28 | you meant |
---|
0:06:31 | and the when the size of the structure immense increase |
---|
0:06:36 | more and more objects more dark objects disappear we can see also the shape of |
---|
0:06:44 | objects also chant |
---|
0:06:47 | for the corn the police like tango |
---|
0:06:51 | column is now a change to be lumped |
---|
0:07:09 | well um |
---|
0:07:16 | okay i was despising |
---|
0:07:18 | that's take |
---|
0:07:20 | one early example on how the a hyperspectral image to see the results when the |
---|
0:07:31 | size of structure demands increase we see a lot of |
---|
0:07:37 | bright |
---|
0:07:38 | small objects disappear i don't know what you can see these uh small disappear uh |
---|
0:07:47 | this one this one this one and the a small bright small objects disappear us |
---|
0:07:54 | the size of structuring elements increase |
---|
0:07:58 | but we can also see that the shape of this object at change |
---|
0:08:04 | you see this lack tango object now the corner it's long it long |
---|
0:08:14 | in order to proceed the shape of objects while always use morphological profiles these reconstructions |
---|
0:08:24 | this christmas that the all the site the shape of objects can be preceded by |
---|
0:08:29 | the then this know the constructions |
---|
0:08:33 | we see here this object |
---|
0:08:37 | the each of these black tango object it's we'll proceed how well the we see |
---|
0:08:43 | that some small object which are expected to disappear now |
---|
0:08:50 | state present in the image for this small loads point load by load which are |
---|
0:08:59 | disappear in here we solve the constructions but now disappear we can say that this |
---|
0:09:07 | reconstructions the size of object mott proceeds about the remotes proceed the well |
---|
0:09:16 | in order to overcome all these problems morphological profiles this posture become structures are used |
---|
0:09:25 | this is not set we can see us the size of the structure it immense |
---|
0:09:31 | increase |
---|
0:09:33 | more and more small point |
---|
0:09:35 | objects disappear corresponding to i |
---|
0:09:39 | structuring elements size at the same time we can proceed the age of the object |
---|
0:09:45 | button then result the constructions |
---|
0:09:55 | so we propose |
---|
0:09:57 | uh extended morphological profiles this posture become could decode reconstructions to combine spectral and the |
---|
0:10:04 | spatial information of the hyperspectral image |
---|
0:10:09 | this is our proposed by set we take features extracted by pca as example to |
---|
0:10:16 | build extended morphological profiles responsive reconstructions we first use the feature extractions delete use that |
---|
0:10:25 | dimensionality and the led dependency of the hyperspectral data and then repute we use all |
---|
0:10:33 | morphological openings and closings this partial reconstruction is to build morphological profiles on each extracted |
---|
0:10:42 | features |
---|
0:10:43 | and then we stack all these features as well |
---|
0:10:48 | as important as the svm classifier and then finally we got the classification maps |
---|
0:10:57 | that's take one example take two extracted features as pca to see we take two |
---|
0:11:06 | we take two openings and the closings and the we start these features us |
---|
0:11:17 | uh so but uh this is extended morphological profiles and the we use this as |
---|
0:11:22 | input to do the classification |
---|
0:11:26 | we can see by increasing the size of the structure immense more and more objects |
---|
0:11:32 | are removed |
---|
0:11:34 | and the extended morphological profiles county the spatial information such as the size of objects |
---|
0:11:41 | in the image |
---|
0:11:44 | the morphological profiles contain |
---|
0:11:47 | yeah O explained this morphological profiles contain four openings and the full closings |
---|
0:11:55 | and the this morphological profile with partial reconstruction is was proposed in to solve i |
---|
0:12:02 | it's with peak |
---|
0:12:06 | last row some experimental results to see if wishes all of all proposed method |
---|
0:12:14 | the four states we use it's problem |
---|
0:12:18 | university of probably i italian the states have one hundred and the fifteen bands |
---|
0:12:25 | some bands are removed due to the noise leaving one hundred and sleep and on |
---|
0:12:30 | the left this is the false color image it's |
---|
0:12:35 | this image has high spatial resolution is one point three meters topic cells the online |
---|
0:12:42 | classes in this image |
---|
0:12:47 | we select ten labels impose a class landlady to change svm classifier based |
---|
0:12:55 | a pf kernels |
---|
0:12:58 | and then test on the testing on the test set |
---|
0:13:02 | morphological features up builds on the first poll can open a simple components |
---|
0:13:11 | and the lease is the results we used original hyperspectral image to do the classification |
---|
0:13:17 | and this results it's |
---|
0:13:19 | we use that we the morphological profile with no reconstruction is and this is the |
---|
0:13:25 | morphological profile with reconstruction and this is all proposed which posture because structures |
---|
0:13:32 | phone to be zero maps we can see that's this more formal logical features the |
---|
0:13:39 | classification maps it's much smoother than results what spatial information |
---|
0:13:47 | this is noise |
---|
0:13:48 | and the compared to the morphological risk we construct use morphological profiles result reconstructions and |
---|
0:13:57 | vese posture because actors a much and smooth then this one |
---|
0:14:03 | and the full morphological profiles with no reconstruction is the shape of some object a |
---|
0:14:10 | chant |
---|
0:14:12 | for this model the corner here it's no longer |
---|
0:14:17 | and we can also compare the overall classification accuracy here |
---|
0:14:23 | here this the morphological features the overall accuracy can improve a lot even then what |
---|
0:14:31 | kind that sense and that all proposed mess that what's the best results |
---|
0:14:38 | and the beacon please see the size and shape but the then some existing methods |
---|
0:14:47 | let's see some results when the change simple size increase |
---|
0:14:52 | the extended morphological profiles built on different features because C |
---|
0:14:59 | the morphological features proposed you on the |
---|
0:15:03 | pretty simple component features |
---|
0:15:06 | as the simple size increase all proposed it can be morphological profiles this posture reconstructions |
---|
0:15:13 | got busted results |
---|
0:15:16 | and |
---|
0:15:18 | the morphological profiles you on kernel pretty simple components and lies is always get the |
---|
0:15:24 | possibly start |
---|
0:15:27 | in some cases it is button want to use the morphological profiles with reconstructions for |
---|
0:15:33 | here this morphological profile |
---|
0:15:38 | we speak objections cannot will be will be was then this no reconstructions this is |
---|
0:15:46 | due to the old reconstructions of morphological profiles by reconstruction |
---|
0:15:57 | let's see and not the plates that's a list that's it's prone city centre a |
---|
0:16:03 | party a |
---|
0:16:04 | it's has one hundred fifteen bands and some bad a limited due to the noise |
---|
0:16:09 | leaving one hundred and two bands on the left side it's the false color image |
---|
0:16:15 | is it also have high spatial resolution is one point three meters but excels yeah |
---|
0:16:22 | nine different classes in this image |
---|
0:16:27 | we select pen labels impose a class blunder the problem this training set |
---|
0:16:35 | and the land passed on the touch screen set of this |
---|
0:16:41 | morphological features appear on the top the first can cut kernel principal components |
---|
0:16:50 | and the this is the result from people and that's that for a while what |
---|
0:16:53 | dates we just use only the original hyperspectral data to do the classification and this |
---|
0:17:00 | is no reconstruction this is this reconstruction and this is with all proposed |
---|
0:17:06 | posture reconstructions because shape on that a little overall classification accuracy all postmaster what's the |
---|
0:17:15 | highest accuracy |
---|
0:17:22 | the see the performance when the change simple size increases |
---|
0:17:28 | but the changes impose size increase all propose |
---|
0:17:33 | extended morphological profiles with partial reconstructions got busted results and that this is always you |
---|
0:17:41 | on the kernel principal components |
---|
0:17:49 | so that's see the conclusion and the future work |
---|
0:17:53 | uh in this presentation we can see that's kernel pretty simple component and vice is |
---|
0:17:59 | more efficient to extracted features for comes construct in uh extended morphological profiles and the |
---|
0:18:09 | extended morphological profiles with partial reconstruction is more efficient competing T then lowest be snow |
---|
0:18:17 | constructions and the miss cost reconstructions |
---|
0:18:21 | and the extended morphological profiles responsive reconstructions europe on kernel pretty simple components are always |
---|
0:18:29 | perform fast especially before there's more training symbols |
---|
0:18:34 | all future work maybe explore different structure meant structure element because now in this presentation |
---|
0:18:43 | and the in my complex paper we just use the circle structure element and that |
---|
0:18:51 | we can see also this kernel principal component we need more features |
---|
0:18:57 | so it is but that that's it's great eight high dimensionality of the extended morphological |
---|
0:19:05 | profiles this posture reconstructions |
---|
0:19:10 | uh that's my presentation thank you |
---|
0:19:28 | yeah |
---|
0:19:37 | uh yes uh i prepared to try the linear structure elements which may efficient for |
---|
0:19:44 | the loans for the streets and the label of the streets they are long lines |
---|
0:19:49 | and the for the labels also like this one i think it's uh maybe it's |
---|
0:19:54 | nice feature to improve the classification accuracy |
---|
0:19:58 | yes |
---|
0:20:37 | a yes uh yes we can use some computers before the classification this may have |
---|
0:20:44 | some i think maybe some the same result because you smoother the original hyperspectral image |
---|
0:20:53 | and the this different scalars you also can smooth the image |
---|
0:20:59 | in different |
---|
0:21:00 | uh how this yeah maybe can see that that's what if devoicing possibly denoising the |
---|
0:21:07 | hyperspectral image and that we use that you noise image to do the classification |
---|
0:21:12 | yes |
---|
0:21:51 | so secure for suggestion i also read some level as follow your group yes uh |
---|
0:21:57 | the extent it's attribute profiles yes it's but uh and the is we because see |
---|
0:22:04 | that because uh it's sometimes it not related with this structure you meant so in |
---|
0:22:10 | may be our future work we can see the we use the attributes uh profiles |
---|
0:22:15 | to improve the classification accuracy that's good for test thank you |
---|
0:22:36 | uh |
---|
0:22:43 | yeah uh we normally we selected this according to the queue mel other human let |
---|
0:22:50 | steve |
---|
0:22:51 | cumulative but like the lions |
---|
0:22:54 | for pca for kernel principal components the first we extracted features three percent ninety nine |
---|
0:23:01 | percent of the two melodies alliance so for kind of pca how the for stole |
---|
0:23:09 | components a reasonable kernel principal components people are sent ninety nine |
---|
0:23:15 | a sense of the cumulative alliance so we just depend on the |
---|
0:23:22 | yeah i don't different dates that's we have different |
---|
0:23:26 | the different number of the extracted features we use yeah |
---|
0:23:33 | yeah |
---|
0:23:36 | yeah that's this also creates a new challenge because you also the dimensionality of the |
---|
0:23:43 | extended morphological profiles will be higher than other than principal component like this |
---|
0:24:00 | i |
---|