JPH0691609B2 - Linear array sensor abnormality detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to automatic detection of anomaly detectors of a linear array sensor, and in particular enables highly accurate anomaly detection by obtaining an image average value for each detector. A linear array sensor abnormality detection method suitable for

[Conventional technology]

Conventionally, spike-like noise is detected due to abnormality of the detector,
The method of interpolating the image data of the detector from the surrounding image data is discussed in Yasuoka et al., National Pollution Research Institute Report No. 39 (1983), pages 119 to 128.

That is, let r be the coordinate in the direction perpendicular to the image scanning line, N (r) be the neighborhood section, and f (r) be the original image obtained as the intensity distribution of the cross section perpendicular to the scanning line. An image is represented by a first-order polynomial, f (r) = α + βr + ε (r) (1) in the neighborhood section, and the coefficients α and β are obtained by the least square method.
That is, the estimation formula is Becomes The determination as to whether the image data of r = 0 is spike noise is It is realized by a hypothesis test of whether or not it deviates from the 95% confidence interval of, that is, by comparison with the seven distribution table. Once, r = 0
If it is possible to detect that the image data, that is, the scanning line data, is abnormal, the scanning line data is discarded and interpolated with the surrounding image data.

[Problems to be solved by the invention]

Since the above-mentioned conventional technique also considers that an abnormality in the scan data occurs in the middle of the image, the abnormality detection processing is performed at each point on the image. There is a problem in that, depending on the difference in the values), it may be determined that an abnormality is found at one place and normal at another place, and therefore the accuracy of detecting abnormal scan data may be reduced.

An object of the present invention is to solve the above problems and provide a highly accurate linear array sensor abnormality detection method.

[Means for solving problems]

The above object is achieved by calculating an image average value for each detector of the linear array sensor image and detecting spike noise on the average value for each detector.

[Action]

The image average value for each detector is represented by F (r) with respect to the detector coordinate, that is, the coordinate r in the vertical direction of the scanning line. If a detector abnormality occurs at r = 0 and a fixed value b is output regardless of the imaging target, the detector average image value F
(0) is also b. By the way, the abnormality detector coordinate r
The variance of the image intensity in the vicinity N (r) of = 0 is smaller than the variance of the original image one-line intensity f (r) with respect to the average value F (r) for each detector. In a special case where there is no correlation between adjacent pixels of the original image, if the number of lines for calculating the average value is M, the image intensity variance in the neighborhood N (r) decreases in proportion to 1 / M. Even when there is a correlation between adjacent pixels of the original image, when the number of lines M is sufficiently large, it also decreases by 1 / M. In other words, original image 1
The average value F (r) for each detector changes more smoothly than the line intensity f (r). Therefore, the accuracy of the abnormality determination of the image average value b for the abnormality detector is significantly improved. According to equation (3), the denominator is Therefore, even if the difference between the estimated value from the surrounding data and b is small, the abnormality determination can be accurately performed.

〔Example〕

An embodiment of the present invention will be described below with reference to FIGS. The linear array sensor 15 consisting of a large number of detectors mounted on the earth observation satellite 1 observes the ground surface 16 line by line with a width of W through the lens system 18 as the satellite moves 17, and transmits image data to the ground. To do. The antenna 2 sends the received image data to the high-speed data recording system 3, and the recording system 3 records the image data on the high-density magnetic tape 4. The image data once recorded on the high-density magnetic tape 4 is read from the magnetic tape device 5 one scan line at a low speed in accordance with the capability of the image processor 6. Image data fj of the r pixel read in the jth line
(R) is added by the processor 6 with the cumulative image Sj _-1 (r) up to the j-1th line read from the cumulative image data buffer 7 according to the following equation, and Sj (r) = Sj _-1 (R) + fj (r) (4) It is written in the accumulated image data buffer 7 again. The image data read from the magnetic tape device 5 is written to the disk device 8 separately from the accumulating process. After the predetermined image line M is read, spike noise detecting means 9
Reads the accumulated image data S _M (r) for each detector from the accumulated image data buffer 7, divides by M to obtain the average value f (r), and then performs spike noise detection processing. FIG. 3 shows a flowchart of this processing procedure. The process of FIG. 3 is performed for each detector with the number r. In box 19, the coefficient of the least-squares linear approximation formula for the average image intensity f (i), iεN (r) {indicated by a mark} in the neighborhood N (r) with the number r as shown in FIG. Ask for. Here, n is the number of data points in the neighborhood N (r). In box 20, the square sum Se (r) of near-order errors from the linear approximation formula is obtained for the neighborhood N (r). box
In 21, the image average value f (r) and its estimated value for the detector with the detector number r in question Difference from Is the approximation error variance of the neighborhood Calculate the value compared with. In box 22, the 95% confidence interval is compared with seven distributions, and if larger, r is output as an anomaly detector in the anomaly detector number file 10 in a box 23. Anomaly data interpolation processor 11 is an anomaly detector file
The abnormality detector number r is read from 10 and the following interpolation processing is performed.

fj (r) = {fj (r-1) + fj (r + 1)} / 2 (5) The interpolation process result fj (r) is overwritten with abnormal data on the magnetic disk 8. The image data of the magnetic disk 8 is then subjected to conventional correction processing such as correction of minute sensitivity unevenness of the detector and correction of geometric distortion in the image complementation processing system 12,
The data is output from the magnetic tape device 13 to the magnetic tape 14.

Further, the present invention can be implemented not only as an image processing method of the satellite ground station equipped with the linear array sensor described above, but also applicable to detector abnormality detection of other linear array type image sensor such as an X-ray CT or fax machine. . With an X-ray CT system,
Since the abnormality of a small number of detectors gradually appears on the reconstructed image as the deterioration of the image quality of the entire image, it is necessary to process the original sensor image data, automatically detect the abnormality detector data, and perform the interpolation process according to the present invention. The effect of improving the image quality is extremely large. Further, in the fax machine, even without using a highly accurate abnormality detection algorithm as in the above embodiment, the absolute value of the difference in average value between the detectors is determined, and whether or not there is a difference exceeding the set threshold value on both sides is checked. If determined, the present invention can be easily implemented.

〔The invention's effect〕

According to the present invention, the automatic detection of the abnormality detector of the linear array sensor can be performed with high accuracy, and since the abnormality determination processing is performed by processing the one-dimensional data, the speed is higher than the determination processing of the entire surface of the two-dimensional image. effective.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of an embodiment of the present invention, FIG. 2 is a diagram showing a state of ground surface scanning by a satellite-mounted linear array sensor, and FIG. 3 is an abnormality detector number from an image average value for each detector. FIG. 4 is a flowchart of a detection processing procedure, and FIG. 4 is a conceptual diagram of hypothesis detection of abnormal data based on a linear approximation formula of neighborhood data.

Claims (1)

[Claims] 1. An image processing system comprising a linear array sensor having a plurality of detectors, an input device for an image picked up by the sensors, an image processing device, and an image output device. Abnormal linear array sensor characterized by calculating an average value, detecting spike-like noise in the average value based on the calculated average value, and correcting the detector output corresponding to the detected spike-like noise Detection method.