Apllication assessment of multi-level Remote Sensing and multi source informationin theinvestigation of forest resources

Apllication assessment of multi-level Remote Sensing and multi source informationin theinvestigation of forest resources

You Xianxiang, Cui Saihua, Peng Fengju
Beijing Forestry University

Abstract
Image of remote is sensing is the means of transmitting messages Different levels and different information sources have different admit ability to transmit information in this paper the aspects of the extraction of different information geometric accuracy classification accuracy and applied potential were studied including space remote sensing (MSS, TM, SPOT, national land resource satellite), different kinds of air photos (color infrared photograph, panchromatic photograph), with different scales (1/130,000, 1/70,000, 1/30,000, 1/20,000, 1/8700). The application ability of the above items in the investigation of forest resources was evaluated.

Classification method and accuracy analysis about MSS TM SPOT and Chinese Territory Satellite (CTS)
The investigation of forest resources lies mainly in checking the quantity and distribution feature of the resources and mastering types of forest vegetation spices of trees types of stand and state of growth. The various data of quantity and quality of land suitable for forest are also to be collected for this purpose enhancement filter and information extraction of different vegetation indices were used and then optimum band combination was conducted for national land resources land resources image only optical amplification treatment was carried out.

Table 1 shows the automatic classification results of MSS combined image after enhancement the classification results of TM combined image after the processing of equalized D-value and the second principal component and NIR/RED the classification results of SPOTI combined image after the processing of Fourier transformation ratio and enhancement the classification results of SPOT2 combined image after piece wise linear stretching these methods are used because the effect of classification is good.

The interpretation ability of the applicable part image and the color image taken with national land resources satellite of the applicable part image can be known by comparison with TM and SPOT images in the following table.

Table. 1 Method and Accuracy (MSS, TM, SPOT) Classification results.

Classification accuracy(%)	Classification
Classification accuracy(%)	Pinus tabulea- forais	larix principis- rupprechtii	betula platyphylla & populus davidina	quercus liaotun- gensis	brush	grass land	Fields	wild land	resi- dentail points	body of water	average accuracy %
MSS754 *	92.5	37.5	85.6	80.8	43.7	78.0	69.0	95.0			77.9
TM453 *	92.0	86.0	77.0	90.0	86.0	43.9	77.0	90.0			84.0
SPOT1 *	91.3		98.4	97.5	91.5	100.0	81.3	88.8			93.2
	Conifer		robinia pesudaocaia	valley shelter forest				bare wild land
SPOT2	89.2		81.5	100.0	90.0	79.2	81.0	86.7	54.6	100	84.7

*Experiment sites of MSS,TM,SPOT1 are in the same region, experiment site of SPOT2 in hilly region.

Comparison between the interpretation properties of (CTS) and that of TM image

LAND TYPE	TM	CTS
FIELD AND VALLEY FIELD	Block distribution of a large place of field can be clearly distinguished color of crops in the field can be distinguished sore clearly than that of a place of national land but texture of the former is not so clear as that of the letter	Border of the field can be distinguished the block of the field that can be distinguishedis much smaller than TM the sate of growth of crops can be distinguished the advantage of block empty land and two row field and two row field shelter forest net meatiness not easily distinguished on TM images is even greater than that pf SPOT satellite in terms of distinction of filed shelter forest net.
SLOPING FIELD	It can be distinguished only from color and cannot be directly interpreted there is no border line of sloping field it is not easy to distinguish from sparse grass land.	The border line of sloping field can be distinguished there is distinct texture structure but add pieces of small blocks are not easy to difference.
SPECIES	On normal combination image some species are not easy to distinguish but most of the on enhancement image can be different such as populus the robinia and brush small they are easy to mis- differentiate	On color photo of national land resource satellite species are more easy to different mainly in the aspects of color texture and law of distribution brush is easily distinguished from tree but on black white picture be higher than TM image easy to differentiate.
CanOpy Density	On TM color combination image denseness can be different from sparseness but separable grades are limited	In the aspects of texture structure homogenous degree distribution and color canopy density can property be divided in to several grades interpretation of canopy density can be higher than spot image
GRASSLAND	Larger pieces of grassland can easily be distinguished ones can be distinguished only in the aspect shade of color but easy to mis distinguish such as from on color different it is more grassland can be different but it is more difficult to differentia the grades of grassland on color combination image on the photo of national	Sealer area of grassland can be distinguished to texture structure and color grades of grassland can also be interpreted.
TREES PLANTED ON FOUR SIDES	Tree planted on four sides are not easy to distinguish expect for this planted incites and towns	Tree planted on four sides of smaller villages can be distinguished trees planted on four sides of large villages are easy to distinguish.
ROAD	The main artery is visible it is difficult to distinguish the branch line the main artery in hilly area is easy to break.	Both the main artery and branch line can be distinguish the cart road can also distinguished
CITIES AND TOWN RESIDENTIAL POINTS	Average cities and town can be distinguished but streets in countries aren't easy to distinguish small villages are not easy distinguish either.	The main streets cities and towns can be distinguished most small villages can be distinguished the interpretation property is higher than SPOT image
SMALL RIVER BODY OF WATER	The average small branch valley is not easy distinguish the small reservoir in the gusty some times is not easy to distinguish either.	The average small branch valley can be distinguish the small reservoir can also be distinguished.

Others such as the distinction of geomorphic feature and corrosion valley and the interpretation property of color image taken with national land resources satellite are also higher than TM image but in the aspects of providing image processing color mixture combination of wavebands and geometric accuracy the advantages of TM image are much greater than national land resources Satellite.

From above comparison it can be seen that MSS classification accuracy is the lowest TM lies between MSS and SPOT in classification accuracy in the 3 methods the classification accuracy of all residential points in the hilly area is low forest land in the complete forest area has higher classification effect sloping field terraced field bare land grassland are easy to mis-differentiate Yet the interpretation ability in the middle part of national land resource satellite image is higher than SPOT and TM.

The above experiment proves that because different information sources have different spatial resolving ability different spectrum resolving ability and different wavelength scope and band combinations there is different classification accuracy.

Contrast of Geometric accuracy
When the forest resource investigation of an average scope is made with the scale 1/250.000--1/50,000 map the optical correction satellite image controlled by a topographic map in general can satisfy the requirement of area accuracy the contrast of the following geometric accuracy is conducted on the requirement of area accuracy the contrast of the following geometric accuracy is conducted on the image only by optical by correction on the premise that the displacement of image pints and the error of geometric accuracy analysis of image line segment measurement are taken as the item of accuracy assessment in the analysis.(Table 2-5)

Table 2 Accuracy table of line segment length measurement in hilly area on TM image.

Order number	1:5ten thousand length of topographic(m)	1:4ten thousand length of TM image(m)	D-value	D-value square	Measurement accuracy
1	3190	3196.00	6.00	47.33	99.7%
2	3550	3577.8	27.89	772.84	99.2%
3	4540	4596.96	33.94	1091.64	99.3%
4	5545	5589.2	44.20	1953.46	99.2%
5	3895	3054.0	40.52	1641.87	98%
Total				5507.33
Mf= ± 33.188

Table 3 Accuracy table of line segment length measurement in hilly area on SPOT image

Order number	1:5ten thousand length of topographic	12,5ten thousand length of TM image	D-value	D-value square	Accuracy
1	6000	5953.2	46.8	2190.3	99.2%
2	2190	2182.0	27.89	63.60	99.6%
3	2725	2755.3	33.94	918.1	99.9%
4	2950	2964.3	44.20	294.5	99.5%
5	450	442.3	40.52	51.8	99.4%
Total				3424.5
Mf = ± 26.17

Table 4 Accuracy analysis of area measurement of TM image and 1"30.000 color
infrared aerial photograph transfer photo area

Order number	Land type	1:30.000 color infrared transfer area(ha)	1:40,000 in transfer area(ha)	D-value	Accuracy%
1	Robinia pesudoacaia	2.75	3.04	0.29	98
2	Conifer	12.0	12.17	0.17	98.6
3	Bare land	33.0	32.0	0.10	99.7
4	Brush	5.0	4.5	0.50	90.0
5	Deciduous forest	19.5	19.2	0.30	98.5
6	Field	37.0	37.2	0.30	99.27
Average accuracy		96.0%

Table 5 Accuracy analysis of measured area on SPOT image and
1/250,000 orthophotograph area

Order number	1:25,000measured on orthophotograph	1:12,500area on SPOT image	Area D-value	Accuracy %
1	8.75	8.7	0.05	99.5
2	9.5	9.4	0.1	98.9
3	3.2	3.27	0.07	97.8
4	31.5	31.20	0.30	99.0
5	2.1	2.0	0.1	95.27
6	7.0	7.2	0.2	97.27
Average accuracy %		97.37%

Geometric accuracy of national land resource satellite image is measured with two forms 19.4 cm * 23cm and 167cm *19.4 cm the image of 19.4 cm *23cnm are divided into four parts central middle top bottom right top bottom and left top bottom error in central part is the slightest middle top bottom 2.5-7% right top bottom 5---9% left top bottom 3.11% on the ling strip image of 167 cm * 19.4 cm with the central points as the centre of the circle and the radius of 5cm 16 concentric circles of equal space length are made the geometric accuracy of each concentric circle is measured according to the sequence of top middle bottom the result is that error in the central part is about 2.2-2.4% that in the fifth concentric circle is about 4-8% and error in the edge part of the sixteenth concentric circle is maximum higher than 30%.

From the above it can be seen that both TM and SPOT have higher geometric accuracy the geometric accuracy of the national land resource satellite image without correction is lower and that the distribution of error varies with the changes of altitude distortion and topography.

Comparison of the smallest distinction area
The minimum distinction area is related to the clarity degree of land type border and clarity degree of land type border depends on topographic factor and other factors of images processing method broken degree of land and the difference among the reflectance values of land type etc.(table 6).

Table 6The smallest distinction area in experimental region on Tm SPOT image

Image	TM				SPOT
The smallest distinguished area land type	Minimum resolving ability(ha)				Minimum resolving ability(ha)
The smallest distinguished area land type	Good background	Number of pixel	Bad background	Number of pixel	Good background	Number of pixel	Bad background	Number of pixel
Conifer	0.37	4.1	0.45	5.0	0.2	5.0	0.3	7.5
Robinia Pesudoacaia	0.6	6.7	1.0	11.1	0.16	4.0	0.2	5.0
Deciduous forest	0.5	5.6	0.8	8.9	0.2	5.0	0.25	6.3
Brush	0.41	4.6	0.6	6.7	0.2	5.0	0.37	9.3
Valley shelter-forest	More than two rows		More than 3 rows		In 2 rows of background ,one rows of large trees can be distinguished.
Grassland	0.2	0.2	0.3	3.3	9.15	1.7	0.2	5.0
Field	0.4	4.4	0.63	7.0	0.25	6.3	0.4	10.0
Bare land	0.2	2.2	0.6	6.7	0.2	5.0	0.45	11.3

Comparison Between different aerial Photographs and scales
In the research different scales of panchromatic photograph color photograph and color infrared photograph are used.

Table 7 comparison of interpretation accuracy between panchromatic photograph
color infrared aerial photograph of different scales*

Kind of photo Type of interpretation	1:30,000 black-white panchromatic photo	1:30,000 color infrared	1:70,000 color infrared	1:30,000color infrared	1:20,000color infrared
Interpretation accuracy of land type	88%	93%	87%	72%	More than 95%
Interpretation accuracy of vegetation type	50-70%	95-98-100%	75%	76%	97-100%

*Experimental area is selected in the region with less complicated vegetation..
1/87000 color infrared aerial photograph is used in investigation growth of trees in the city number of trees and given grass land index (ratio of vegetation cover to the total measured area) it can determine the growth status of trees in the city determined accuracy of trees number can reach more than 85% determined accuracy of green land index more than 90%.

In view of the above comparison and the requirements of specialty map making in land detailed inventory and the second type survey of forest resources 1/20.000 and 1/30,000 color infrared Photograph or panchromatic photograph is good 1/70,000 color infrared photographic tolerable 1:30.000 color infrared photograph of good quality after enlargement can tolerably satisfy the requirements of land type interpretation in the investigation of forest state growth and pollution color infrared aerial photograph is good of the city 1/5000/1/5000 color infrared photograph is needed As for the forest area survey 1/20,000 1/30,000 color infrared photo is needed About 1/130,000 color infrared aerial photograph can be used in macroscopic survey the effect of its enlargement can satisfying the requirement o land type survey in larger regions .

Of the series of satellite images SPOT is the best but its cost is 7-8 times that of TM MSS image is generally used only in macroscopic survey and cannot met the requirements of land detailed survey and the first type investigation in forest resources inventory However after processing TM image can meet part of requirements of land detailed survey and the second type forest resources survey if part of aerial photograph and ground survey are supplemented it can meet the greater part the requirements in the aspect of land type area SPOT can basically meet the greater part of requirements of land detailed survey and the second type investigation in forest resources inventory.