CN1987346A - Method and device for quick high precision positioning light spot image mass center - Google Patents

Abstract

The invention discloses a fast and high-precision spot image centroid positioning method, which performs Gaussian convolution operation on the pixel gray value, and judges whether the pixel gray value after the Gaussian convolution operation is greater than a preset threshold, and if so, the current Mark the pixels, identify the facula they belong to, and calculate the cumulative value of the product of the gray value of the current pixel and the coordinate value and the product of the gray value and coordinate value of all pixels of the same facula that have been processed. The gray value of the current pixel is the same as the processed one. The accumulated value of the gray value of all pixels of the spot, save the accumulated value, otherwise, mark the current pixel as a background pixel and process it; after processing the entire output image, calculate the accumulated value of the gray value and the cumulative value of the coordinate value and the accumulated value of the gray value The quotient of the value, and the calculation result is output as the coordinate value of the center of mass of the spot image. The invention also discloses a centroid positioning device at the same time. The invention can improve the data processing speed and anti-noise ability in the spot image centroid positioning, and can process multiple spot images.

Description

Fast and high-precision spot image centroid location method and device

technical field

The invention relates to machine vision detection technology, in particular to a fast and high-precision spot image centroid positioning method and device.

Background technique

The spot image is common image information in machine vision and pattern recognition, and the center of the spot is the feature of the spot image. Spot center is widely used in target tracking in machine vision, feature point extraction of high-precision three-dimensional measurement in visual inspection, positioning of laser spot center in deep space laser communication for space applications, star point positioning of star sensor of attitude measurement components, Sun spot positioning for sun sensors.

At present, the localization methods for the spot center can be divided into two categories: grayscale-based localization methods and edge-based localization methods. Among them, the grayscale-based positioning method generally uses the gray-scale distribution information of the target spot image for positioning, and can use the centroid method, surface fitting method, etc.; the edge-based positioning method generally uses the edge shape information of the target spot image for positioning. , including: edge circle (ellipse) fitting, Hough transform, etc.

The grayscale-based positioning method has higher accuracy than the edge-based positioning method. Usually, the grayscale-based surface fitting method uses a Gaussian surface to fit the grayscale distribution of the target spot image, but the commonly used two-dimensional Gaussian surface The calculation of the function is more complicated. Therefore, the centroid method has become the most widely used positioning method because of its relatively simple implementation and high positioning accuracy. There are some improved forms of the centroid method, mainly including the centroid method with a threshold and the square-weighted centroid method. Among them, the centroid method with a threshold is equivalent to subtracting the original image from the background threshold, and calculating the centroid for the pixels greater than the threshold in the original image. ; The square weighted centroid method uses the square of the gray value instead of the gray value as the weight. This method highlights the influence of the larger gray value pixels that are closer to the center on the center position.

In the prior art, in space applications that require high real-time visual dynamic tracking, measurement, and miniaturization requirements, spot center positioning is to process images with a large amount of data, and these processing processes have great parallelism. Including operation parallelism, image parallelism, neighborhood parallelism, pixel bit parallelism, etc. However, at present, the spot center positioning method is mainly realized by software on the computer. Because the software implementation is serially executed according to instructions, the spot center positioning becomes the bottleneck of image data preprocessing. Therefore, in terms of real-time spot center positioning, the Jet Propulsion Laboratory (JPL) of the United States proposed a window-based centroid positioning device, which is realized by an analog circuit and embedded in an image sensor chip. This centroid positioning device can simultaneously locate the image centroid of multiple windows, but because the device mainly uses analog circuits and uses a window-based data processing method, there are the following defects in implementation:

1) The setting of the spot processing window is inflexible, and the window cannot be set too large, otherwise, more than two spots that may exist in the window will be treated as one spot, and the result obtained will be wrong;

2) The approximate position and range of the light spot in the image must be known in advance before setting the window;

3) Limited by the processing speed and transmission speed, too many windows cannot be set, so when the number of spots in the image is large, all the spots in the image cannot be obtained;

4) Due to the implementation of analog circuits, this method is sensitive to noise, and the presence of noise will cause a large error in positioning.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a fast and high-precision spot image centroid positioning method, which can improve the data processing speed and anti-noise ability in spot image centroid positioning, and can perform any number of spot images of any size. deal with.

Another object of the present invention is to provide a fast and high-precision spot image centroid positioning device, which can solve the bottleneck problem and noise sensitivity problem of image preprocessing with a large amount of data in spot image centroid positioning, and can detect any number of spots of any size The image is processed.

In order to achieve the above object, technical solution of the present invention is achieved in that way:

A fast and high-precision spot image centroid positioning method, comprising the following steps:

A. Perform a Gaussian convolution operation on the pixel gray value, and judge whether the pixel gray value after the Gaussian convolution operation is greater than a preset threshold, if yes, perform step B, otherwise, perform step C;

B. Mark the currently read pixel, identify the spot to which the current pixel belongs, and calculate the cumulative value of the product of the current pixel gray value and coordinate value and the product of all pixel gray values and coordinate values of the same spot that has been processed, the current pixel The cumulative value of the gray value and the processed gray value of all pixels of the same light spot, save the obtained cumulative value, and execute step D;

C. Mark the current pixel as a background pixel, and judge whether to adjust the stored data of the light spot to which the current pixel belongs, if yes, adjust the stored data of the light spot to which the current pixel belongs, otherwise perform step D;

D. Judging whether to process the entire output image, if it is not processed, return to step A, if it is processed, calculate the quotient of the cumulative value of the product of the gray value of each spot obtained in step B and the coordinate value and the cumulative value of the gray value , and output the obtained quotient as the centroid coordinate value of each spot image.

Wherein, in the step B, the step of merging equivalent marks in the same spot is further included while marking. Before performing the Gaussian convolution operation, it further includes: reading the gray value of the current pixel, and buffering the gray value of the currently read pixel.

In the above method, marking the current pixel in step B further includes:

B11, judging whether the mark value of the left pixel of the current pixel is zero, if not zero, then mark the current pixel as the mark value of the left pixel, and perform step B13, otherwise, perform step B12;

B12. Determine whether the mark value of the pixel above the current pixel is not zero, if so, mark the current pixel as the mark value of the above pixel, and perform step B13, otherwise, mark the current pixel as a new mark value, and update the new mark value ;

B13. Assign the current pixel marker value to the corresponding marker parameter in the left marker parameter and the upper marker parameter group.

In the above method, the merging of equivalent labels in the same spot further includes:

B21, judge the mark value of the pixel on the left side of the current pixel and the pixel above, if they are all zero, then set the equivalent mark parameter corresponding to the current pixel as a new equivalent mark value, update the new equivalent mark value, and perform step B22; is not zero, and the two are not equal, then the number of merged marks will be +1, and step B22 will be executed;

B22. Determine whether the number of merged marks is 1, if the number of merged marks is 1, then merge the equivalent marks of the pixel on the left side of the current pixel into the equivalent marks of the pixel above the current pixel, and update the new equivalent mark value to the previous new etc. value mark value; if the combined mark number is not 1, then execute step B23;

B23. Determine whether the equivalent mark value of the pixel on the left side of the current pixel is equal to the equivalent mark value of the pixel above the current pixel, if not, merge the equivalent data, and merge the equivalent mark value of the pixel above the current pixel into the left side of the current pixel Pixel equivalence marks, if equal, do not process.

In the above method, the step C further includes: clearing the upper mark parameter group and the left mark parameter to zero; the judgment in step C is: judge whether the mark value of the pixel to the left of the current pixel is greater than zero, and if so, adjust the value of the current pixel. The storage data of the light spot, otherwise no adjustment; the adjustment is: accumulating the value of the accumulator into the data memory corresponding to the equivalent mark value, and clearing the accumulator.

The present invention also provides a fast and high-precision spot image centroid positioning device, including a Gaussian filter unit, a spot recognition unit, and a spot centroid calculation unit, wherein the Gaussian filter unit is used to perform Gaussian filtering on the gray value of the output image pixel, and Send the pixel gray value processed by Gaussian filtering to the spot centroid calculation unit; the spot recognition unit is used to receive the control signal for spot recognition input by the spot centroid calculation unit, and complete the pixel marking of the spot image; the spot centroid calculation unit, according to The pixel mark value is used to calculate the centroid of different spot images, and the final calculation result is output.

The spot recognition unit further includes: a mark judger, a left mark register, an upper mark register group, a current mark register, and a new mark register; wherein, the mark determiner is used to mark pixels; the left mark register, the upper mark register group , the current flag register, and the new flag register are used to store and provide the flag value of the pixel to the left of the current pixel, the flag value of the pixel above the current pixel, the flag value of the current pixel, and the new flag value to the flag judger.

The spot identification unit also includes: a merge equivalence mark judge, a merge mark register, a new equivalence mark register, and an equivalence mark buffer; wherein, the merge equivalence mark judger is used to identify the equivalence marks in the same spot Merge; equivalence mark cache, for storing the merged equivalent mark value; merge mark register, for storing merge mark value; new equivalence mark register, for providing new equivalence to the merge equivalence mark judger Flag value; left flag register, upper flag register group, and current flag register, which are further used to provide the flag value of the pixel to the left of the current pixel, the flag value of the pixel above the current pixel, and the flag value of the current pixel to the combined equivalent flag judger.

In the above device, the Gaussian filtering unit further includes: an image cache and a Gaussian convolution operation unit, wherein the image cache is used to cache the gray value of the read pixel and provide it to the Gaussian convolution operation unit; the Gaussian convolution operation unit The computing unit is used to complete the Gaussian convolution operation on the cached pixel gray value, and output the calculation result to the spot centroid calculation unit.

The spot centroid calculation unit further includes: a row and column counter for calculating and providing the coordinate value of each pixel point; a threshold comparator for comparing the Gaussian convolution operation pixel gray value output by the Gaussian convolution operation unit and Presetting the threshold, and outputting the comparison result as a control signal; the first and second multipliers are respectively used to calculate the product of the pixel gray value and the x coordinate value, and the product of the pixel gray value and the y coordinate value; the first and The second adder is used to calculate the cumulative value of the product of the pixel gray value and the x coordinate value, the cumulative value of the product of the pixel gray value and the y coordinate value, and send the obtained cumulative value to the first and second data respectively memory storage; the third adder is used to calculate the cumulative value of the gray value of the pixel, and sends the obtained cumulative value to the third data memory for storage; the first, second and third data memory are respectively used to store the gray value of the pixel The cumulative value of the product of the degree value and the x coordinate value, the cumulative value of the product of the pixel gray value and the y coordinate value, and the cumulative value of the pixel gray value; the first divider and the second divider are used to calculate the pixel gray value respectively The quotient of the accumulated value multiplied by the x coordinate value and the accumulated value of the pixel gray value, the quotient of the accumulated value multiplied by the pixel gray value and the y coordinate value and the accumulated value of the pixel gray value.

The fast and high-precision spot image centroid positioning method and device provided by the present invention perform Gaussian filter processing on the pixel gray value of the output image, and simultaneously mark and calculate each pixel of the output image, so that a or more than one spot image for automatic identification and processing. The present invention has the following advantages:

1) The present invention adopts the Gaussian weighted centroid positioning method, first performs Gaussian convolution operation on the output image data, completes Gaussian filtering, and then performs spot recognition, thereby improving the anti-noise ability of the method and device, and realizing high-precision position.

2) The present invention marks and processes each pixel of the entire output image instead of using a window form, so any number of light spots in the image can be identified and processed, and the size and shape of the light spots are not limited.

3) When the present invention initially marks the light spot, since there may be more than one equivalent mark for the same light spot, the image data of the same light spot is stored in more than one data buffer. The equivalent marks are combined, and the image data belonging to the same light spot is cached in the same data buffer by merging the equivalent marks and compressing the equivalent mark values, which can greatly save data storage space.

4) Since the present invention realizes marking pixels, merging equivalent marks, and cumulative calculation of pixels in parallel, and adopts an FPGA hardware device for real-time realization, it solves the bottleneck problem of image preprocessing with a large amount of data, so that the highest data update rate can reach 30MHz, which can realize real-time centroid extraction.

Description of drawings

Fig. 1 is a flow chart of a specific embodiment of the spot image centroid location method of the present invention;

Fig. 2 is a flow chart of performing pixel marking in the process shown in Fig. 1;

Fig. 3 is a schematic diagram of a marked spot image;

Fig. 4 is the flow chart of merging equivalence marking in the flow process shown in Fig. 1;

Fig. 5 is a schematic diagram of the spot image after equivalent labeling and merging of the spot image shown in Fig. 3;

FIG. 6 is a schematic diagram of the composition and structure of a specific embodiment of the spot image centroid positioning device of the present invention.

Detailed ways

The basic idea of the present invention is to firstly perform Gaussian filter processing on the pixel gray value of the output image, and then perform marking and calculation processing on each pixel of the output image at the same time, so that one or more than one spot images can be automatically identified and processing.

Here, the marking and calculation process specifically includes: comparing each output pixel, marking each spot pixel, and combining equivalent marks for different pixels in the same spot when necessary, so as to ensure that each pixel of the same spot Each pixel is given the same mark, and different spot pixels have different marks; while marking and merging, the gray value and coordinate value product of the same marked pixel are accumulated and the gray value is accumulated; after the output of the entire image data is completed, The cumulative value of the product of the gray value and the coordinate value of the same marked pixel is divided by the cumulative value of the gray value to obtain the centroid positioning coordinates of each light spot. In this way, fast and high-precision spot centroid positioning for multi-spots with unlimited sizes and shapes can be realized.

It can be seen from the centroid positioning process of the prior art that noise has a great influence on the positioning accuracy. Therefore, the Gaussian weighted centroid positioning method is adopted in the present invention, that is, the gray value of the original image pixel is not used for calculation when the centroid is positioned, but Gaussian filtering is performed on the gray value of the original image pixel by formula (1), and then the gray value of the original image pixel after Gaussian filtering is used for calculation.

$\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\overset{\frac{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\overset{\frac{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In formula (1), F(x, y) represents the gray value of the output image data, I(x, y) represents the gray value of the output image data after Gaussian convolution processing, and g(i, j) represents the Gaussian filter coefficient .

Fig. 1 shows the processing process of a specific embodiment of the method for locating the centroid of the spot image of the present invention, referring to Fig. 1, the method for locating the centroid of the spot image of the present embodiment includes the following processing steps:

Step 101: Read the current pixel gray value, and cache the currently read pixel gray value.

Here, the cache pixel gray value is generally determined according to the size of the Gaussian convolution template, that is, the amount of row data of the output image. Subsequent processing is carried out after the 7th row.

Steps 102-103: Perform Gaussian convolution operation on the cached grayscale value, and compare the pixel grayscale value obtained after the Gaussian convolution operation with the preset threshold value, and check whether the pixel grayscale value after the operation is greater than the set threshold value , if yes, it means that the current pixel is a light spot, go to step 104 to identify the light spot; otherwise, it means that the current pixel is a background, go to step 107.

Here, the Gaussian convolution operation is also determined according to the Gaussian convolution template, for example: for a 7×7 Gaussian convolution template, the 7 rows and 7 columns of the output image are processed each time, and the processing order is usually output image The starting point shall prevail, from left to right and from top to bottom; the specific Gaussian convolution operation adopts the calculation method of formula (1). In fact, the specific implementation method only needs to achieve the calculation purpose of formula (1). The threshold is usually determined according to the contrast between the grayscale of the output image itself and the background. Generally, the smaller the spot contrast, the lower the threshold setting; the larger the spot contrast, the higher the threshold setting.

Step 104: Mark the currently read pixel and identify the light spot to which the current pixel belongs.

Among them, background pixels can be marked with zero, and non-background pixels can be marked with non-zero values. Of course, in practical applications, the background pixels can also be marked with other values, correspondingly, the non-background pixels can be marked with non-background pixel marking values, as long as the background and non-background and different light spots can be distinguished. For the convenience of calculation and marking, zero and positive integers are generally used as optional marking values. Of course, negative integers, decimals, etc. can also be used. The following steps are described by taking background pixels marked as zero and non-background pixels marked as non-zero positive integers as an example.

The specific marking process of each pixel is shown in Figure 2, including the following steps:

Steps 104-104b: Determine whether the marked value of the pixel to the left of the current pixel is zero, if not, mark the current pixel as the marked value of the left pixel, execute step 104f, and if it is zero, execute step 104c.

Steps 104c-104e: Determine whether the mark value of the pixel above the current pixel is zero, if not, mark the current pixel as the mark value of the pixel above, execute step 104f, if it is zero, mark the current pixel as a new mark value , and update the new tag value.

Here, the new flag value can be stored in a special register to provide a new flag value for the pixel, and the new flag value can be updated in different ways, as long as the new flag value provided each time is not repeated. For example: after each use of the new tag value, add 1 to the new tag value and save it again for the next pixel tag use.

Step 104f: Assign the current pixel label value to the corresponding label parameter in the left label parameter and the upper label parameter group, so as to be used for labeling the next pixel and the next row of pixels.

Here, the upper flag parameter group may be stored in the register, and the left flag parameter may be stored in the register. Among them, the left marker parameter is a marker value, which is set to zero during initialization, and the upper marker parameter group is used to save a group of marker parameter values. An array can be used, and each marker in this group corresponds to a pixel, for example: a row has 10 pixels, the upper marker parameter group is a marker group consisting of 10 markers, each marker corresponds to a pixel in the row, and the initial values of the marker parameters in this group are all zero. Correspondingly, when assigning a value, assign the tag value of the current pixel to the tag parameter in the upper tag parameter group corresponding to the current pixel, for example: there are 10 pixels in a line, the upper tag parameter group includes 10 tag parameters, and the current pixel is The 5th pixel of the row to which it belongs, then the assignment refers to assigning the tag value of the current pixel to the 5th tag parameter in the upper tag parameter group. When judging, the tag value of the pixel above the current pixel is also judged by finding the tag parameter corresponding to the serial number of the current pixel in the upper tag parameter group.

Steps 104a-104f are a pixel marking process, and each pixel in the output image can be marked by repeating steps 104a-104f. For example: for the pixel in row 2 and column 4 in Figure 3, first judge the mark value of the pixel to the left of the current pixel, because it is equal to zero, so continue to judge the mark value of the pixel above the current pixel, which is also equal to zero, then mark the current pixel as new tag value, and update the new tag value. Another example: for the pixel in row 2 and column 5 in FIG. 3 , first judge the mark value of the pixel to the left of the current pixel. Since it is equal to 2, the current pixel is directly marked as 2.

Step 105: Merge equivalent markers in the same light spot.

Fig. 3 is a schematic diagram of an image marked by the method in Fig. 2. The area covered by the shadow in Fig. 3 is a light spot, and there are four light spots in Fig. 3 . It can be seen from Fig. 3 that for the same spot, there may be multiple different marks, which are equivalent to the same spot. Therefore, in order to unify all the marks in the same spot, the present invention adopts the method shown in Fig. 4 The process of merging the equivalent marks gives each spot the same equivalent mark value. When the background mark is zero, the equivalent mark value is also a positive integer starting from 1. The specific process of merging equivalent marks is shown in Figure 4, including:

Steps 105a-105c: Determine the tag values of the left pixel and the upper pixel of the current pixel, if they are all zero, set the equivalent tag parameter corresponding to the current pixel as the new equivalent tag value, update the new equivalent tag value, and execute step 105d ; If both are not zero, and the two are not equal, it means that the marks of the two are equivalent, and the number of merged marks will be +1, and step 105d will be executed.

Here, the new equivalent tag value can be stored in a special register to provide a new equivalent tag value for the pixel, and the new equivalent tag value can be updated in different ways, as long as the new equivalent tag value provided each time is guaranteed Tag values are not repeated. For example: after each use of the new equivalent tag value, add 1 to the new equivalent tag value and save it again for the next pixel tag use. The number of merge marks is used to record the number of equivalent marks that need to be merged, and the value of the number of merge marks can be stored by a register, and the value of the finally obtained equivalent marks can be stored in a special register.

Steps 105d-105h: Determine whether the number of merged marks is equal to 1, if so, merge the equivalent marks of the pixel to the left of the current pixel into the equivalent marks of the pixel above the current pixel, and update the new equivalent mark to the previous new equivalent tag value. If the update new equivalent tag value in step 105b is to add 1 to the new equivalent tag value each time, then, here, updating the new equivalent tag value to the previous new equivalent tag value is to subtract 1 from the current new equivalent tag value .

Since the new equivalent tag value is updated to the previous new equivalent tag value in the process of merging the equivalent tag, the range of the equivalent tag value is compressed, and the equivalent tag value is the address corresponding to the data memory, and the new equivalent tag Compression of ranges saves data storage units considerably. As shown in Figure 3, the data storage unit used when the equivalent mark compression is not performed is 19, and one pixel mark corresponds to one data storage unit, and most of the storage units are empty and useless, and the equivalent mark is performed The compressed spot image is shown in Figure 5, only 4 data storage units are needed.

If the number of merged marks is not equal to 1, it is further judged whether the equivalent mark value of the pixel to the left of the current pixel is equal to the equivalent mark value of the pixel above the current pixel, and if they are not equal, merge the equivalent data and the merge of the equivalent marks . Here, the equivalent data refers to the data in the storage space corresponding to the equivalent tag. Specifically: merge the data in the data memory space corresponding to the equivalent mark of the upper pixel into the storage space corresponding to the equivalent mark of the left pixel, clear the data memory space corresponding to the equivalent mark of the upper pixel, and at the same time clear the data of the current pixel The equivalent marks of the upper pixel are merged into the equivalent marks of the left pixel of the current pixel, and if they are equal, they are not processed.

The result of the merged image shown in Figure 3 is shown in Figure 5. It should be noted that the mark on each pixel in Figure 5 is actually the address of the data memory where the spot image data to which the pixel belongs is finally stored, for example: Figure 5 The spot at the upper left of the center is marked as 1, which means that the spot image data of this spot is stored in the data memory equivalently marked as 1.

In practical applications, if all pixels in a spot have adopted the same tag value, then there is no need to perform equivalent tag merging; or, if the reduction of storage space is not considered, this step can also be omitted, so step 105 is optional.

Step 106: Accumulate the product of the gray value of the current pixel and the coordinate value and the accumulated value of the product of the gray value of all pixels of the same spot that has been processed and the coordinate value, and combine the gray value of the current pixel with all the gray values of the same spot that have been processed Accumulate the accumulated gray value of the pixel, save the obtained accumulated value, and execute step 110 .

In the present invention, the above-mentioned steps 104, 105 and 106 are implemented in parallel for each pixel, so that the processing speed can be greatly improved.

Step 107: mark the current pixel as a background pixel. In this embodiment, mark the current pixel as zero, and clear the upper mark parameter group and the left mark parameter to zero.

Here, the definitions of the upper label parameter group and the left label parameter are exactly the same as those described in step 104f.

Steps 108-109: Determine whether the mark value of the pixel to the left of the current pixel is greater than zero, if not, directly execute step 110; if yes, adjust the stored data of the light spot to which the current pixel belongs, the specific operation is: add the value of the accumulator to equivalent tag value in data memory and clears the accumulator.

Step 110: Judging whether the entire output image has been processed, if yes, execute step 111, otherwise return to step 101. Here, it may be determined whether the output image is processed completely according to whether the end flag of the current output image is read.

Step 111: According to the formula (2), divide the accumulated value of the product of the gray value and the coordinate value calculated in step 106 by the accumulated value of the gray value, and output the obtained quotient as the coordinate value of the center of mass of the spot image.

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ,</span>$ ${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

In formula (2), I(x, y) represents the gray value of the output image data after Gaussian convolution processing; x ₀ and y ₀ are the x and y coordinate values of the centroid of the spot image.

In order to realize the above method, the present invention proposes a corresponding spot image centroid positioning device. As shown in FIG. Wherein, the spot identification unit 61 is used to receive the control signal for spot identification input from the threshold comparator 632 in the spot centroid calculation unit 63, and complete the pixel labeling of the spot image and the combination of equivalent labels of the same spot. The spot identification unit 61 further includes a mark judger 611, a merge equivalent mark judger 612, a left mark register 613, an upper mark register group 614, a current mark register 615, a new mark register 616, a merge mark register 617, a new equivalent mark register 618. The equivalence tag buffer 619.

Wherein, the marking determiner 611 is used to mark the pixels, and the specific marking process adopts the process shown in FIG. A marker for the current pixel.

The merging equivalence mark judging unit 612 is used for merging the equivalent marks in the same spot. The specific merging process adopts the process shown in FIG. The value tag register 618 completes the combination of equivalent tags for different pixels of the same light spot and saves the value of the equivalent tag in the equivalent tag buffer 619 .

Current tag register 615, left tag register 613, upper tag register group 614, new tag register 616, merge tag register 617, new equivalent tag register 618, equivalent tag register 619 are respectively used for storing and sending to tag judger 611, Merge equivalent flag determiner 612 provides the flag value of the current pixel, the flag value of the pixel to the left of the current pixel, the flag value of the pixel above the current pixel, the new flag value, the merge flag value, the new equivalent flag value and the final equivalent flag value . Among them, the upper tag register group 614 and the equivalent tag buffer 619 are used to store a set of tag parameter values, for example: a tag value of a row of pixels, and only one tag value is stored in the remaining registers. The equivalence tag buffer 619 also provides the equivalent tag of each spot after merging as an address to the data memory 635a, 635b and 635c in the spot centroid calculation unit 63, so that the image data of each spot is finally stored in the combined In the data memory corresponding to the equivalence tag.

If the equivalence mark merge is not performed, the merge equivalence mark determiner 612, the merge mark register 617, the new equivalence mark register 618 and the equivalence mark register 619 can be omitted.

The Gaussian filter unit 62 is used to perform Gaussian filtering on the grayscale value of the output image pixel, and send the pixel grayscale value processed by the Gaussian filter to the spot centroid calculation unit 63; the Gaussian filter unit 62 further includes an image buffer 621 and a Gaussian convolution The computing unit 622, wherein the image cache 621 is used to cache the gray value of the read pixel and provide it to the Gaussian convolution computing unit 622; the Gaussian convolution computing unit 622 is used to complete the Gaussian convolution on the cached pixel gray value and output the calculation result to the threshold comparator 632, multipliers 633a and 633b, and adder 634c in the spot centroid calculation unit 63.

The spot centroid calculation unit 63 is used to calculate the centroid of the spot image, and output the final calculation result. The spot centroid calculation unit 63 further includes: a row and column counter 631, which is used to calculate and provide the coordinate value of each pixel, and input the x and y coordinate values of each pixel to the multipliers 633a and 633b respectively; the threshold comparator 632 uses After receiving the Gaussian convolution operation pixel gray value output by the Gaussian convolution operation unit 622 and the preset threshold value input separately, the two are compared, and the comparison result is sent to the mark judger 611 as a control signal, combined Equivalence mark determiner 612, adders 634a, 634b and 634c.

The spot centroid calculation unit 63 further includes: multipliers 633a and 633b, adders 634a, 634b and 634c, data memory 635a, 635b and 635c, and dividers 636a and 636b. Among them, the multipliers 633a and 633b receive the Gaussian-filtered pixel gray value output by the Gaussian convolution operation unit 622 and the x and y coordinate values input by the row and column counters, and output the product of the pixel gray value and the coordinate value; the adder 634a , 634b and 634c receive the output result of the threshold comparator 632 and its own accumulation result, and respectively receive the output result of the multipliers 633a, 633b and the Gaussian convolution operation unit 622 for accumulation operation. In fact, the multiplier 633a and the adder 634a are used to calculate the cumulative value of the product of the x coordinate value of all pixels of the same light spot and the pixel gray value, and store the calculation result in the data memory 635a; the multiplier 633b and the adder 634b are used to Calculate the cumulative value of the product of the y-coordinate values of all pixels of the same spot and the gray value of the pixel, and store the calculation result in the data memory 635b; the adder 634c is used to calculate the cumulative value of the gray value of all pixels of the same spot, and store the calculation result Stored in the data storage 635c. The dividers 636a and 636b are respectively used to calculate the quotient of the cumulative value of the product of the x coordinate value and the pixel gray value and the cumulative value of the pixel gray value, the cumulative value of the product of the y coordinate value and the pixel gray value and the cumulative value of the pixel gray value The quotient of the spot to get the x-coordinate value and y-coordinate value of the centroid of the spot. The multipliers 633a and 633b, the adders 634a, 634b and 634c, the data storages 635a, 635b and 635c, and the dividers 636a and 636b are used to complete the calculation of formula (2).

The device for locating the spot image centroid of the present invention can be realized by using a field programmable gate array (FPGA, Field programmable gate array) or an application specific integrated circuit (ASIC, Application Specific Integrated Circuit).

When performing spot centroid positioning on the output image, the device shown in FIG. 6 reads the current pixel gray value from the current output image, and caches the currently read pixel gray value in the image buffer 621. After that, the image buffer The pixel gray value in 621 is sent to the Gaussian convolution operation unit 622 for Gaussian convolution operation to complete the Gaussian filter; the pixel gray value through the Gaussian convolution operation is input into the threshold comparator 632, which is compared with the pre-set value input separately. Set the threshold to compare, then determine whether to carry out spot recognition according to the comparison result, if yes, then the threshold comparison result is input to the mark judger 611 and the merged equivalent mark judger 612 in the spot recognition unit 61 as a control signal, and start to identify the spot Marking of pixels and merging of equivalent marks, the specific marking and merging process passes through mark judger 611, merge equivalent mark judger 612, left mark register 613, upper mark register group 614, current mark register 615, new mark register 616 , merge mark register 617, new equivalence mark register 618 and the cooperation between equivalence mark register 619, finish by the flow process shown in Fig. 2 and Fig. 4; Simultaneously, by multiplier 633a and 633b, adder 634a, 634b And 634c, data memory 635a, 635b and 635c complete the accumulation of the product of the pixel gray value and the coordinate value and the accumulation and storage of the pixel gray value; after the entire output image is determined to be processed, the spot centroid is calculated by the divider 636a and 636b The x and y coordinate values of . Wherein, the x and y coordinate values of each pixel are provided by the row and column counter 631 .

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (11)

1. A fast high-precision spot image centroid positioning method is characterized by comprising the following steps:

A. performing Gaussian convolution operation on the pixel gray value, and judging whether the pixel gray value subjected to the Gaussian convolution operation is larger than a preset threshold value or not, if so, executing the step B, otherwise, executing the step C;

B. marking the current read pixel, identifying the light spot to which the current pixel belongs, calculating the product of the gray value and the coordinate value of the current pixel and the accumulated value of the products of the gray value and the coordinate value of all pixels of the same processed light spot, storing the obtained accumulated value, and executing the step D;

C. marking the current pixel as a background pixel, and judging whether to adjust the storage data of the light spot to which the current pixel belongs, if so, adjusting the storage data of the light spot to which the current pixel belongs, otherwise, executing the step D;

D. and C, judging whether the whole output image is processed or not, if not, returning to the step A, and if so, calculating the quotient of the accumulated value of the product of the gray value and the coordinate value of each light spot obtained in the step B and the accumulated value of the gray value, and outputting the obtained quotient as the coordinate value of the centroid of each light spot image.

2. The method for locating the centroid of the spot image according to claim 1, wherein the step B further comprises a step of merging equivalent marks in the same spot while marking.

3. The method for positioning the centroid of the spot image according to claim 1 or 2, wherein the performing the gaussian convolution operation further comprises: and reading the gray value of the current pixel, and caching the gray value of the current pixel.

4. The method for locating the centroid of the light spot image according to claim 1 or 2, wherein the marking the current pixel in step B further comprises:

b11, judging whether the marking value of the left pixel of the current pixel is zero, if not, marking the current pixel as the marking value of the left pixel, and executing the step B13, otherwise, executing the step B12;

b12, judging whether the marking value of the pixel above the current pixel is not zero, if so, marking the current pixel as the marking value of the pixel above, executing the step B13, otherwise, marking the current pixel as a new marking value, and updating the new marking value;

and B13, assigning the current pixel mark value to the corresponding mark parameter in the left mark parameter and the upper mark parameter group.

5. The spot image centroid localization method according to claim 2, wherein said merging of medium valence marks in the same spot further comprises:

b21, judging the marking values of the left pixel and the upper pixel of the current pixel, if the marking values are zero, setting the equivalent marking parameter corresponding to the current pixel as a new equivalent marking value, updating the new equivalent marking value, and executing the step B22; if both are not zero and the two are not equal, merging the mark number +1, and executing the step B22;

b22, judging whether the number of the merged marks is 1, if so, merging the equivalent marks of the pixels on the left side of the current pixel into the equivalent marks of the pixels above the current pixel, and updating the new equivalent mark value to be the previous new equivalent mark value; if the merging flag number is not 1, executing step B23;

b23, judging whether the equivalent mark value of the left pixel of the current pixel is equal to the equivalent mark value of the upper pixel of the current pixel, if not, merging the equivalent data, merging the equivalent mark of the upper pixel of the current pixel into the equivalent mark of the left pixel of the current pixel, and if so, not processing.

6. The spot image centroid positioning method according to claim 1 or 2, wherein the step C further comprises: clearing the upper marking parameter group and the left marking parameter;

and step C, judging as follows: judging whether the marking value of the left pixel of the current pixel is larger than zero, if so, adjusting the stored data of the light spot to which the current pixel belongs, otherwise, not adjusting;

the adjustment is as follows: and accumulating the value of the accumulator into the data memory corresponding to the equivalent mark value, and clearing the accumulator.

7. A fast high-precision light spot image centroid positioning device is characterized by comprising a Gaussian filter unit, a light spot identification unit and a light spot centroid calculation unit, wherein,

the Gaussian filtering unit is used for carrying out Gaussian filtering on the gray value of the pixel of the output image and sending the pixel gray value subjected to the Gaussian filtering processing to the light spot centroid calculating unit;

the light spot identification unit is used for receiving a control signal for light spot identification input by the light spot mass center calculation unit and finishing pixel marking of a light spot image;

and the light spot centroid calculating unit is used for calculating the centroids of different light spot images according to the pixel marking values and outputting the final calculation result.

8. The spot image centroid locating device according to claim 7, wherein said spot identification unit further comprises: the system comprises a mark judger, a left mark register, an upper mark register group, a current mark register and a new mark register; wherein,

a marking judger for marking the pixels;

the left marking register, the upper marking register group, the current marking register and the new marking register are used for storing and providing the marking value of the left pixel of the current pixel, the marking value of the pixel above the current pixel, the marking value of the current pixel and the new marking value for the marking judger.

9. The spot image centroid locating device according to claim 8, wherein said spot identification unit further comprises: a merging equivalent mark judger, a merging mark register, a new equivalent mark register and an equivalent mark buffer; wherein,

a merging equivalence mark judger for merging equivalence marks in the same light spot;

the equivalent mark buffer is used for storing the combined equivalent mark value;

a merge flag register for storing a merge flag value;

a new equivalence mark register for providing a new equivalence mark value to the merging equivalence mark judger;

the left marking register, the upper marking register group and the current marking register are further used for providing the marking value of the left pixel of the current pixel, the marking value of the upper pixel of the current pixel and the marking value of the current pixel for the merging equivalent marking judger.

10. The spot image centroid locating device according to any one of claims 7 to 9, wherein said gaussian filtering unit further comprises: an image buffer and a gaussian convolution operation unit, wherein,

the image buffer is used for buffering the gray value of the read pixel and providing the gray value to the Gaussian convolution operation unit;

and the Gaussian convolution operation unit is used for completing Gaussian convolution operation on the cached pixel gray value and outputting the calculation result to the light spot mass center calculation unit.

11. The spot image centroid positioning device according to any one of claims 7 to 9, wherein the spot centroid calculating unit further comprises:

the row-column counter is used for calculating and providing the coordinate value of each pixel point;

the threshold comparator is used for comparing the pixel gray value subjected to the Gaussian convolution operation and output by the Gaussian convolution operation unit with a preset threshold, and outputting a comparison result as a control signal;

the first multiplier and the second multiplier are respectively used for calculating the product of the pixel gray value and the x coordinate value and the product of the pixel gray value and the y coordinate value;

the first adder and the second adder are respectively used for calculating the accumulated value of the product of the pixel gray value and the x coordinate value and the accumulated value of the product of the pixel gray value and the y coordinate value and respectively sending the obtained accumulated values to the first data memory and the second data memory for storage;

the third adder is used for calculating the accumulated value of the pixel gray value and sending the obtained accumulated value to the third data memory for storage;

the first data memory, the second data memory and the third data memory are respectively used for storing the accumulated value of the product of the pixel gray value and the x coordinate value, the accumulated value of the product of the pixel gray value and the y coordinate value and the accumulated value of the pixel gray value;

the first divider and the second divider are respectively used for calculating the quotient of the pixel gray value and the x coordinate value multiplication accumulated value and the pixel gray value accumulated value and the quotient of the pixel gray value and the y coordinate value multiplication accumulated value and the pixel gray value accumulated value.

Images