JPH0827832B2 - Method and apparatus for classifying and evaluating image data - Google Patents

Description

The present invention relates to a method and apparatus for classifying and evaluating image data, and more particularly, to the accuracy of a measurement target area classified into a plurality of attributes, that is, for each attribute. The present invention relates to a method and apparatus for classifying and classifying image data suitable for evaluating accuracy and change from image data.

[Prior Art] Conventionally, as a map search device, a device described in Japanese Patent Laid-Open No. 61-62176 is known, but as represented by a land use classification map, etc. In order to identify the secular change and error of the area in which the land is classified into multiple attributes, for example, vegetation characteristics such as forests, fields, waste land, urban characteristics such as urban areas, etc. The accuracy of the area for each attribute was evaluated by visual interpretation using photographs.

[Problems to be solved by the invention]

The above-mentioned conventional method requires manual evaluation when evaluating the certainty of each attribute area, and thus quantitative evaluation is difficult. In addition, even when performing land-based evaluation over a wide area, it was necessary to rely on visual interpretation of photographs, etc. by hand, and it was not possible to accurately identify the secular change or error of the area for each attribute. .

An object of the present invention is to provide an image data classification / evaluation method and apparatus capable of evaluating the certainty of a region for each attribute from measurement data obtained by imaging a region to be measured.

[Means for solving problems]

In order to achieve the above-mentioned object, the present invention obtains measurement data by sequentially capturing a region to be measured with images having different spectral characteristics, and according to classification data obtained by dividing the region to be measured into a plurality of attributes. A method for classifying image data, which classifies the measurement data, calculates the correlation characteristics of each attribute based on the classified measurement data, and evaluates the likelihood of each attribute of the classification data from the calculated correlation characteristics. It was adopted. Still further, according to the present invention, there is provided a first memory for storing classification data in which a measurement target area is divided into a plurality of attributes, and measurement data obtained by sequentially capturing the measurement target area with images having different spectral characteristics. And a third memory for storing a judgment reference value regarding the correlation characteristic for each attribute, and the measurement data of the second memory for each attribute according to the classification data of the first memory. Measurement data classification means, correlation characteristic calculation means for calculating the correlation characteristic of each attribute from the classified measurement data, and the calculated correlation characteristic and the judgment reference value of the third memory for comparison. The image data classification / evaluation apparatus includes a classification data evaluation unit that evaluates the likelihood of each attribute and an evaluation result output unit that outputs the evaluation result of the probability of each attribute.

[Action]

The region to be measured is actually measured and divided into a plurality of attributes to generate classification data. Further, the region to be measured is sequentially imaged with images having different spectral characteristics to obtain measurement data.
The measurement data is classified for each attribute according to the classification data, and the correlation characteristic of each attribute is calculated based on the classified measurement data. The calculated correlation characteristic is compared with the determination reference value for the correlation characteristic to evaluate the certainty of each attribute of the classification data, and the evaluation result is output.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG.

In FIG. 1, the data file 10 is a data file.
Calculator 2 with 10A and 10B via input / output interface
0 is connected, and the computer 20 is connected to the output device 30 via the input / output interface. Data file
In 10A, as shown in Fig. 3 (a), the area to be measured is divided into attributes such as forests and urban areas based on the maps obtained through surveying and field surveys of land use classification maps. The classified data is stored. That is, the areas A1 and A2 are stored as forests in the Betru data, and the areas B1, B2, and B3 are stored as the urban areas in vector data.

The data file B stores measurement data obtained by sequentially capturing the regions to be measured corresponding to the land use classification map with images having different spectral characteristics. The measurement data is, for example, data obtained based on image information captured by a communication satellite. As shown in (b) and (c) of FIG. 3, this measurement data was observed in the wavelength band of band 1 which is well sensitive to the urban area as a spectral characteristic in the image information observed by the communication satellite. It is obtained from image information and image information observed in the wavelength band of band 2 which is sensitive to forests. That is, the image data observed in the wavelength band of band 1 is the image data in which the brightness level of the city area is higher than that of other attributes. Then, each measurement data is stored in association with the position information.

The computer 20 has a CPU, a ROM, a RAM, a memory, etc., and the memory constitutes a third memory for storing a judgment reference value concerning the correlation characteristic for each attribute, and the CPU etc. functions as a measurement data classification means and a correlation characteristic calculation means. It constitutes a classification data evaluation means. The output of the classification data evaluation means is output to the output device 30. The output device 30 is composed of a display device 30A and a printing device 30B as an evaluation result output means.

This embodiment has the above configuration, and its operation
It will be described with reference to the drawings.

When the classification data of the data file 10A serving as the first memory and the measurement data of the data file 10B serving as the second memory are input to the computer 20, the measurement data classifying means sorts the measurement data according to the attributes. are categorized. Then, the correlation characteristic calculation means performs a process of calculating the correlation characteristic of each attribute from the classified measurement data. As the calculation processing of this correlation characteristic, as shown in FIG. 5, the classification region i for each attribute is extracted, the number of pixels Bi of the classification region i is calculated, and the average mi of the brightness levels of the classification region i is calculated. , The variance σi of the luminance level of the classification region i is equal to or higher than the average mi is calculated. These calculations are performed for all areas for each attribute. Then, the calculated data is stored in the memory as the correlation characteristic table.

From the correlation characteristic calculation process shown in FIG. 5, the correlation characteristic regarding the area such as the forest or the urban area is calculated. For example, when the image information as shown in FIG. 3 (b) is obtained as an image regarding a city area, the correlation characteristic as shown in FIG. 4 (b) is obtained. Further, when the image shown in FIG. 3 (c) is obtained as the image regarding the forest, the correlation characteristic as shown in FIG. 4 (a) is obtained. That is, since the brightness levels of the areas B1 and B2 are high but the brightness level of the area B3 is low among the areas B1, B2 and B3 in the urban area, the brightness level of each area is plotted as shown in FIG. Such characteristics can be obtained. Also, when plotting the brightness levels of the forest areas A1 and A2, since the areas A1 and A2 are both observed in the wavelength band of band 2, the brightness level is high,
The characteristic result as shown in FIG. 4 (a) is obtained.

When the correlation characteristic is calculated, the classification data evaluation means compares the correlation characteristic with the determination reference value to evaluate the certainty of each characteristic of the classification data. This evaluation method can be performed, for example, from the state of the variance σi of each classification region in the correlation characteristics. For example, for urban areas, as shown in FIG. 4 (b), the distribution of the brightness levels in areas B1, B2 and B3 is dispersed, so the classification data for urban areas is not accurate. Can be evaluated. That is, the area B3 is evaluated as having changed from the attribute of the city area to another attribute. On the other hand, regarding the forest, as shown in (a) of FIG. 4, it is assumed that the distribution state of the brightness levels related to the areas A1 and A2 is concentrated in one place and is included in the judgment reference value of the variance σi. It is evaluated that the data has accuracy. Then, these evaluation results are classified according to whether each attribute has certainty or not, and are stored in a memory as an evaluation result table for each attribute. Then, the contents of the evaluation result table stored in the memory are output to the output device 30 for the evaluation result output processing for converting into the display signal and the print signal. The display signal and the print signal are respectively supplied to the output device 30 in association with the position information of the land use classification map, and the evaluation result for each attribute is displayed as an image on the screen of the display device 30. For example, a region in which the evaluation result is not accurate is displayed in red. Further, the printing device 30B prints the quantitative data obtained by the calculation processing of the correlation characteristic.

As described above, in this embodiment, since the reliability of the land use classification map can be evaluated based on the image data, the extent to which the land use classification map does not need to be surveyed or surveyed It can be evaluated whether it has certainty. In addition, if the image data of the same area observed by the Landsat satellite for 16 days is used as the measurement data, it is possible to reliably evaluate the secular change and error of the land use classification map.

〔The invention's effect〕

According to the present invention, it is possible to evaluate the certainty of the classification data related to the measurement target based on the image data, and thus it is possible to identify the change or error in the classification data without actually measuring the measurement target.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG.
FIG. 3 is a diagram for explaining the operation of the device shown in FIG. 3, FIG. 3 is a diagram for explaining the structure of the land use classification map and the observation image, and FIG.
FIG. 5 is a correlation characteristic diagram of the areas A and B, and FIG. 5 is a flow chart for explaining the correlation characteristic calculation processing. 10A, 10B ... Data file, 20 ... Calculator, 30 ... Output device, 30A ... Display device, 30B ... Printing device.

Claims (2)

[Claims] 1. A measurement target region is sequentially imaged with images having different spectral characteristics to obtain measurement data, and the measurement data is classified according to classification data in which the measurement target region is divided into a plurality of attributes. A method of classifying and evaluating image data, characterized in that the correlation characteristic of each attribute is calculated based on the measured data, and the certainty of each attribute of the classification data is evaluated from the calculated correlation characteristic. 2. A first memory for storing classification data in which a measurement target region is divided into a plurality of attributes, and measurement data obtained by sequentially capturing the measurement target region with images having different spectral characteristics. A second memory, a third memory for storing a judgment reference value relating to the correlation characteristic for each attribute, and a measurement for classifying the measurement data of the second memory for each attribute according to the classification data of the first memory. Data classification means, correlation characteristic calculation means for calculating the correlation characteristics of each attribute from the classified measurement data, and the calculated correlation characteristics and the judgment reference value of the third memory to compare the attributes of the classification data. An image data classification and evaluation device comprising: a classification data evaluation unit that evaluates the likelihood of each attribute; and an evaluation result output unit that outputs an evaluation result of the likelihood of each attribute.