CN1710596A - Automatic recognition method of computer jigsaw puzzle based on machine vision - Google Patents

Abstract

Image information collection is carried out for picture mosaic pieces disturbed optionally by camera above mosaic pieces working area. Through a collection card, information is transferred to computer for pretreatment. Using method of edge detection and threshold segmentation determines valid mosaic pieces, as well as picks up location information and angle information of mosaic pieces on the working area. Using location information calculates center of mosaic piece, and using angle information rotates mosaic piece to correct position. The, matching between mosaic piece and template till identifying all mosaic pieces is completed and image is recovered finally. Features are high accuracy, good reliability and fast speed.

Description

Computer picture-arrangement automatic identifying method based on machine vision

Technical field

The present invention relates to a kind of computer picture-arrangement automatic identifying method, obtain image, adopt computer vision technique to realize the automatic identification and the coupling of picture mosaic, belong to Flame Image Process and automatic identification technology field according to image capture device based on machine vision.

Background technology

Along with the continuous progress of world today's computer technology, robot vision has become in the fields such as manufacturing industry, check, document analysis, medical diagnosis and military affairs inalienable part in various intelligence, the autonomous system in conjunction with motion control.Because its importance, some advanced countries classify computer vision research as economy and science are had extensive influence science and the great basic problem of engineering.

Automatically identification is that the core in the robot vision technology also is a difficult point wherein, has obtained utilization widely.Such as automatic identification of literal, car plate and people's face or the like.Its algorithm quantity is many, with strong points, will select suitable automatic identifying method for use for specific recognition objective.

In traditional picture arrangement game, after the number of jigsaw puzzle pieces reaches certain number, can make the difficulty of picture mosaic increase greatly, often relatively working of repetition carried out in a slice picture mosaic meeting, reduced efficient widely.The appearance of automatic identification technology can well address this problem.According to patent retrieval, do not find the similar patent that automatic identification technology is applied to picture mosaic.

Chinese invention patent " automatic face-recognizing digital video system " (application number 01128827.2) has been introduced a kind of method of the people's of identification face, discerns hair, forehead, eyes, nose, face successively according to Markov model.Can discern people's face fast, have characteristics such as discrimination height, speed are fast, reliable operation.If but the method is applied in the identification of picture mosaic, then there is the shortcoming of accuracy of identification deficiency.

Summary of the invention

The objective of the invention is at the deficiencies in the prior art, develop a kind of computer picture-arrangement automatic identifying method, can adopt the method for machine vision that the jigsaw puzzle pieces of upsetting is discerned processing automatically, realize the recovery of picture according to template based on machine vision.

In order to realize this purpose, the present invention adopts the method for machine vision, the video camera that utilization is positioned at top, picture mosaic workspace carries out the image information collection to the jigsaw puzzle pieces of arbitrarily upsetting, information is sent to computing machine through image pick-up card and carries out pre-service, the method of utilization rim detection and Threshold Segmentation is determined effective jigsaw puzzle pieces, and each sheet jigsaw puzzle pieces is extracted positional information and the angle information of jigsaw puzzle pieces in the perform region.Utilize the central point of positional information calculation jigsaw puzzle pieces, utilize angle information that jigsaw puzzle pieces is gone back to positive attitude after, mate with template.According to template matching algorithm, choose the best matching result of similarity as recognition result, determine the tram of jigsaw puzzle pieces, until the identification of finishing all jigsaw puzzle pieces, last restored image.

Method of the present invention is undertaken by following concrete steps:

1. gather the image information of picture mosaic workspace by ccd video camera, adopt medium filtering to remove picture noise, gray processing adopts threshold segmentation method with jigsaw puzzle pieces and background segment, finishes the image pre-service.

2. adopt edge detection algorithm that the adjacent image point that belongs to a slice picture mosaic together is carried out mark, and add up the number of pixels that every jigsaw puzzle pieces comprises.When the number of pixels that comprises when certain jigsaw puzzle pieces is less than setting threshold, then it is judged as invalid jigsaw puzzle pieces, does not carry out subsequent treatment.

3. effective jigsaw puzzle pieces is extracted characteristic quantity, travel through the pixel that this jigsaw puzzle pieces comprises, obtain jigsaw puzzle pieces four jiaos coordinate figure up and down.Calculate the center position coordinate (X of jigsaw puzzle pieces thus _Cen, Y _Cen), and jigsaw puzzle pieces is put the angle θ that needs rotation with respect to positive attitude (both sides level promptly).

4. jigsaw puzzle pieces is rotated after the θ angle forwards positive attitude to and put, use and mate based on the template matching algorithm of similarity and all templates of storage, jigsaw puzzle pieces whenever revolves to turn 90 degrees with template and once mates, write down each matching result and the corresponding angle n pi/2 that has turned over, n=0,1,2,3.

5. the comparison match result selects best that time matching result of similarity as recognition result, reads the corresponding angle n pi/2 of coupling this time, n=0,1,2,3, simultaneously this is mated the center point coordinate of the center point coordinate of corresponding templates as jigsaw puzzle pieces, determine the tram of jigsaw puzzle pieces.

6. after all jigsaw puzzle pieces being finished identification, according to the center point coordinate of each jigsaw puzzle pieces and the angle [alpha]=θ+n pi/2 of needs rotation, n=0,1,2,3 carry out emulation, utilize serial ports to the mechanical arm move instruction after emulation finishes, finish automatic picture mosaic by mechanical arm and restore.

Picture-arrangement automatic identifying method of the present invention is put sample at random by 100 spelling figure and is experimentized, and average accuracy rate is 100%, and be no more than 2s the working time of each identification, has accurately and characteristics fast.

The present invention adopts the automatic identifying method of computer vision, and the jigsaw puzzle pieces upset is realized image acquisition, image recognition, thereby finishes the target of image restoration, has characteristics such as accuracy height, good reliability, speed be fast.Designed method of the present invention can send instruction to mechanical arm after the identification of finishing image, finished the recovery of image by mechanical arm.

Description of drawings

Fig. 1 picture mosaic perform region synoptic diagram.

Fig. 2 jigsaw puzzle pieces position feature synoptic diagram.

Embodiment

Below in conjunction with drawings and Examples technical scheme of the present invention is further described.

(80mm * 80mm) is an identifying object to embodiments of the invention, carries out discerning automatically based on the computer picture-arrangement of machine vision with 5 * 5 jigsaw puzzle pieces.Fig. 1 is the synoptic diagram that jigsaw puzzle pieces is put in the perform region.

Automatic identification specific implementation process of the present invention is as follows:

On the basis of arbitrarily upsetting jigsaw puzzle pieces, video camera carries out the image information collection to the picture mosaic perform region, and image information is sent to computing machine through image pick-up card and carries out the image pre-service: image gray processing; Threshold segmentation method is cut apart background and jigsaw puzzle pieces; Medium filtering is removed picture noise.

After the Threshold Segmentation separating background, utilize the rim detection principle that the pixel that belongs to a slice picture mosaic is carried out mark, and add up the number of the pixel that comprises with a slice picture mosaic.Setting threshold is 500, if the number of pixels that comprises of a slice picture mosaic is less than this threshold value, then with this picture mosaic as noise greatly, be considered as invalid jigsaw puzzle pieces and do not carry out follow-up operation.

For effective jigsaw puzzle pieces, travel through the coordinate that pixel that this jigsaw puzzle pieces comprises is obtained these four summits of jigsaw puzzle pieces, i.e. X _Left, Y _Left, X _Right, Y _Right, X _Top, Y _Top, X _Bottom, Y _BottomAs shown in Figure 2, utilize the central point principle to obtain the position (X of jigsaw puzzle pieces central point _Cen, Y _Cen), obtain jigsaw puzzle pieces simultaneously and put the angle θ that needs rotation with respect to positive attitude (both sides level promptly).

Jigsaw puzzle pieces is rotated the θ angle, promptly turn to after positive attitude puts, use is mated based on the template matching algorithm of similarity and all templates of storage, jigsaw puzzle pieces whenever revolves to turn 90 degrees with template and once mates, corresponding every template is carried out four couplings (90 degree, 180 degree, 270 degree, 0 degree) altogether, to carry out mating for four times four similarities that obtain with every template and compare, therefrom select the best matching result of similarity as jigsaw puzzle pieces and this template.Its similarity and 90 number of times of spending that turn over are recorded one 25 * 25 two-dimensional array.

Travel through this two-dimensional array of 25 * 25, choose the best matching result of similarity as recognition result, the template of this recognition result correspondence is the most similar template of jigsaw puzzle pieces therewith.According to the center position of this template obtain matching the center position information that jigsaw puzzle pieces correctly puts (X ' _Cen, Y ' _Cen) and last jigsaw puzzle pieces need angle [alpha]=θ+n pi/2 of rotating, n=0,1,2,3 when correctly putting altogether.

Last according to the center position coordinate of each jigsaw puzzle pieces with respect to angle [alpha]=θ+n pi/2 of correctly putting the needs rotation, n=0,1,2,3 carry out picture mosaic emulation, utilize serial ports to the mechanical arm move instruction after emulation finishes, and finish picture mosaic by mechanical arm and restore.

The automatic identifying method of computer vision picture mosaic proposed by the invention can be realized image acquisition, Flame Image Process, image recognition to the jigsaw puzzle pieces of upsetting, thereby finish the target of image restoration.The experimental result accuracy rate is 100%, and be no more than 2s the working time of each identification, has accurately and characteristics fast.

Claims (1)

1, a kind of computer puzzle automatic recognition method based on machine vision, it is characterized in that comprising the steps: 1) Collect the image information of the jigsaw work area through the CCD camera, use the median filter to remove the image noise, grayscale processing, and use the threshold segmentation method to separate the jigsaw picture and the background to complete the image preprocessing; 2) Use the edge detection algorithm to mark the adjacent pixels belonging to the same piece of puzzle, and count the number of pixels contained in each piece of puzzle, when the number of pixels contained in a certain puzzle is less than the set threshold, it will be If it is judged as an invalid jigsaw piece, no follow-up processing will be performed; 3) Extract the feature quantity for an effective jigsaw piece, traverse the pixels contained in the jigsaw piece, and obtain the coordinate values of the upper, lower, left, and right corners of the jigsaw piece, and thus calculate the coordinates of the center point of the jigsaw piece, and the position of the jigsaw piece relative to the upright posture. Put the angle θ that needs to be rotated; 4) After the jigsaw piece is rotated at an angle of θ to a positive posture, use the template matching algorithm based on similarity to match all the stored templates. The jigsaw piece is matched with the template every time it is rotated 90 degrees, and the matching result of each time is recorded and the corresponding angle that has been turned $\frac{\mathrm{n\π}}{2},\mathrm{no}=\mathrm{0,1,2,3};$ 5) Compare the matching results, select the matching result with the best similarity as the recognition result, and read the angle corresponding to this matching $\frac{\mathrm{n\π}}{2},\mathrm{no}=\mathrm{0,1,2,3},$ At the same time, the center point coordinates of the matching corresponding template are used as the center point coordinates of the jigsaw piece to determine the correct position of the jigsaw piece; 6) After all the jigsaw pieces are identified, according to the coordinates of the center point of each jigsaw piece and the angle to be rotated $\mathrm{\α}=\mathrm{\θ}+\mathrm{n\π}/2,\mathrm{no}=\mathrm{0,1,2,3}$ Carry out simulation. After the simulation is completed, use the serial port to send instructions to the manipulator, and the manipulator will complete the automatic jigsaw restoration.

Images