WO1993001560A1

WO1993001560A1 - Target classifying technique

Info

Publication number: WO1993001560A1
Application number: PCT/US1992/005620
Authority: WO
Inventors: James L. Nettles; Hyun W. Han
Original assignee: Ltv Aerospace And Defense Company
Priority date: 1991-07-02
Filing date: 1992-07-02
Publication date: 1993-01-21

Abstract

A connected component technique enables the analysis of picture elements to identify components having particular characteristics.

Description

TARGET CLASSIFYING TECHNIQUE

The present invention relates generally to a method and apparatus for identifying a target by a vehicle such as a missile. More particularly, the present invention relates to a technique of identifying a target by connected components in an image memory such as a frame buffer.

In scanning laser radar (LADAR) ranging systems, laser pulses are generated and transmitted. The transmitted pulse strikes a target of interest and reflects a portion of the transmitted pulse back to a receiver associated with the transmitter. Such LADAR systems are commonly installed in projectiles such as missiles to determine target type, location, and the range to a target. The time duration between the transmission of a pulse and the reception of the return pulse is convertible into a range.

The two-dimensional image of the various ranges detected is often stored in a two-dimensional memory array for subsequent processing. Conventional connected component analysis involves analyzing adjacent picture elements ("pixels") for common characteristics. A large metal target, such as a tank or armored personnel carrier, will provide similar levels of intensity of the return signal throughout the image of the target, and this level of intensity will be different than for a soft target such as a tree. Thus, if the frame buffer stores pixels indicative of signal return level, adjacent pixel values may be analyzed on a case-by-case basis to map out the extent of the desired target.

Similarly, the frame buffer image may store pixel values indicative of temperature. An exhaust from a target vehicle would show as a hot spot whereas cold metal of the vehicle itself would have a different temperature profile. Such a profile may provide useful information in identifying the target and distinguishing it from other higher or lower value targets.

However, conventional connected component analysis techniques are slow and computationally intensive. These techniques involve analyzing, on a pixel-by-pixel basis, to determine what pixels adjacent each other pixel has the common characteristic, then moving on to the next pixel to repeat the process. This may require many sweeps of the frame buffer in order to identify and classify each of the targets of interest in the entire frame buffer. As used herein, the term "sweep" means reading out the entire contents of a frame buffer for analysis and processing.

Thus, there remains a need for a system that completes that target identifying scheme in one sweep of the frame buffer. Such a system advantageously provides a substantial increase in processing speed.

The present invention overcomes the drawbacks of the prior art by analyzing pixels in a raster scan in one sweep. As each pixel of interest is analyzed, the pixels already analyzed are reviewed; that is the pixels to the left, to the left and above, above, and to the right and above are reviewed to see if they have the same value as the pixel of interest. As used herein, pixels above the pixel of interest refers to these three pixels. An adjacency code is built depending on the results of this review. For example, if the pixel to the left of the pixel of interest has the same value as that pixel, a logical 1 is placed in bit one of the adjacency code; if the pixel above and to the left is the same, a logical 1 is placed is bit two of the adjacency code; and so on. The adjacency code is used as an input to a table to look up an action code. This action code then dictates further processing in analyzing a component of an image.

Figure 1 is a flow diagram that depicts the technique of the present invention.

Figure 1 depicts the technique of the present invention. A two-dimensional image array 10, such as a frame buffer memory, contains an image for analysis. The process of the present invention lends itself to processing the pixels in the image array 10 at video refresh rates and can even be done while the frame buffer is being updated and/or refreshed. The pixels will have 2ⁿ distinct possible values where n is the number of frames in the frame buffer. A value of zero is considered an empty pixel, not part of any component. The process is preferably performed by any appropriately programmed microprocessor.

The process begins by adding a column of zeros before the first column, a row of zeros before the first row, and a column of zero after the last column. These bands of zeros remove any need to treat pixels in the first or last column as special cases. The next step is to initialize a pointer to point to the first row of an image and then to initialize a pointer to point to the first column in that row. Next, in the block marked "prepare adjacency code" the microprocessor determines if the pixel immediately to the left and the three pixels above the pixel of interest, that is above and to the left, directly above, and above and to the right, have the same value as the pixel of interest. This adjacency code is used as an index to an action table to look up an action code. This action code determines what actions are to be taken next and the number of possible action codes is less than number of possible adjacency codes. This alone speeds processing.

If the pixel is empty, the action code will be NULL and processing proceeds to the next pixel. The column is incremented and a check is made to see if the end of the row has been reached. If not, the pixel of interest is not at the end of a row the process returns to the "prepare adjacency code" block. Now, if the action code is no longer NULL, the action code is equal to START. The process will start a new label indicating the beginning of a new target or other feature having adjacent equal value pixels.

After incrementing the column and preparing a new action code, the process returns to the decision process. There are two possibilities: either the new pixel of interest is of equal in value to the adjacent pixels or it is not. If it is equal, then the process prepares an action code equal to ADD. If it is not equal, the process prepares an action code equal to START, to indicate the beginning of a new connected component.

However, there is yet another possibility. It is possible to begin to label a target as two distinct connected regions and later find that, due to contours, the two are actually connected. In this case, the process develops an action code equal to MERGE. In that case, if the labels are not equal, the process merges the labels and if the labels are equal, the process adds that pixel to the existing label.

Once the process reaches the end of a row, it will increment the row counter and repeat the process until it reaches the last pixel in the frame buffer at which point it exits and the microprocess is free to preform other tasks.

As each pixel is added to an existing label in the block labeled "Add Pixel To Existing Label," it provides a variety of data regarding the component identified by that label. For example, the maximum and minimum X and Y values are updated. At the conclusion of the image, the maximum and minimum X and Y values of each identified component are thus provided.

As another example, assume that the pixel values in the two dimensional image array represent colors. The process of the present invention identifies every pixel of a particular color in an image that is connected to pixels of the same color.

Claims

CLAIMS:

1. A method of analyzing pixel values in a pixel array, comprising the steps of:

a. sequentially viewing each of the pixels in the array in a raster scan;

b. determining coincidence of the values of the pixels to the left and above each pixel viewed;

c. preparing an adjacency code based on the determination of coincidence;

d. preparing an action code based on the adjacency code; and

e. developing a label based on the code prepared.

2. A processor programmed to perform the process of claim 1.

3. The process of claim 1 further including the step of storing the codes.

4. The process of claim 1 further including the step of determining if two labels are contiguous and, if they are, merging the two labels.