CN102087745A - Video image sequence flexible target comparison and characteristic extraction algorithm in digital signal processing (DSP) system - Google Patents

Abstract

The invention discloses a video image sequence flexible target comparison and characteristic extraction algorithm in a digital signal processing (DSP) system. In the algorithm, a target is compared and confirmed by a multi-layer and multi-characteristic comparison algorithm by using the characteristics of the flexible target, and target parameters, which can participate in operation, of the flexible target are simultaneously acquired, namely the target comparison of a complex flexible object in the image is solved by using a simple mathematical model to acquire corresponding basic attributes of the target so as to prepare basic data for image analysis processing. The algorithm is applicable to the technical fields with higher requirements on image information, such as video fire disaster sensors, video image sensors and the like.

Description

Comparison of sequence of video images soft objectives and feature extraction algorithm in the dsp system

Technical field

The present invention relates to a kind of dsp system, be specifically related to its algorithm target.

Background technology

Mainly compare and feature extraction in the general image sequence, and generally adopt the basic principle of similitude of profile to compare based on the rigidity target, less to the comparison of flexible article target; The rigidity target changes little for soft objectives, comparison easily, but soft objectives may produce translation and rotary state, is difficult to remove to weigh the state of target in the different images sequence with the profile of rigidity; And if compare with the set of whole impact point, must cause operand too big, be difficult to be implemented in computing in the picture systems such as DSP.

Summary of the invention

The technical problem to be solved in the present invention is the defective that overcomes existing comparison and feature extraction algorithm, a kind of characteristics of utilizing soft objectives are provided, adopt comparison algorithms multi-level, many features to realize relatively affirmation, obtain the target component that can participate in computing of soft objectives simultaneously target.

In order to solve the problems of the technologies described above, the invention provides following technical scheme:

Comparison of sequence of video images soft objectives and feature extraction algorithm in the dsp system, it comprises the steps:

(1), obtain in the image sequence target in adjacent two sub-pictures the most left, the rightest, go up most, the most following four coordinates, form 4 summits of a rectangular box;

(2), obtain in the image sequence peak of target, two coordinates of minimum point in adjacent two sub-pictures; If a plurality of somes equal altitudes are arranged, according to from left to right writing down corresponding two points; And obtain line segment width and position the wideest in the target image;

(3), obtain in the image sequence circularity and the area of target in adjacent two sub-pictures;

(4), in the sequence video image, at first want tracking target and the general movement velocity of confirming target; Then, the degree of overlapping that compares both relevant positions in image with the rectangle frame in the step (1); If degree of overlapping in an empirical value, assert that both are same targets, and carries out next step comparison; If degree of overlapping not in empirical value, confirms that both are not same targets, and abandon comparison;

(5), compare the degree of overlapping of both relevant positions in image with circularity in the step (3) and area; If degree of overlapping in an empirical value, assert that both are same targets, and carries out next step comparison; If degree of overlapping not in empirical value, confirms that both are not same targets, and abandon comparison;

(6), compare the degree of overlapping of both relevant positions in image with the peak in the step (2), minimum point coordinate and the wideest line segment width and position; If degree of overlapping in an empirical value, confirms that both are same targets, comparison finishes; If degree of overlapping not in empirical value, confirms that both are not same targets, the comparison EOP (end of program).

This algorithm utilizes the characteristics of soft objectives, adopts comparison algorithms multi-level, many features to realize relatively affirmation to target, obtains the target component that can participate in computing of soft objectives simultaneously; Promptly utilize simple mathematics model to go to solve the target comparison of complicated flexible article in image, obtain the base attribute of corresponding target, do basic data for image analysis processing and prepare.This algorithm requires than higher technical field image information applicable to video fire hazard sensor, video image sensors etc.

Description of drawings

Accompanying drawing is used to provide further understanding of the present invention, and constitutes the part of instructions, is used from explanation the present invention with embodiments of the invention one, is not construed as limiting the invention.In the accompanying drawings:

Fig. 1 is the logical procedure diagram of algorithm of the present invention.

Claims (1)

1.DSP comparison of sequence of video images soft objectives and feature extraction algorithm in the system, it comprises the steps:

(1), obtain in the image sequence target in adjacent two sub-pictures the most left, the rightest, go up most, the most following four coordinates, form 4 summits of a rectangular box;

(2), obtain in the image sequence peak of target, two coordinates of minimum point in adjacent two sub-pictures; If a plurality of somes equal altitudes are arranged, according to from left to right writing down corresponding two points; And obtain line segment width and position the wideest in the target image;

(3), obtain in the image sequence circularity and the area of target in adjacent two sub-pictures;

(4), in the sequence video image, at first want tracking target and the general movement velocity of confirming target; Then, the degree of overlapping that compares both relevant positions in image with the rectangle frame in the step (1); If degree of overlapping in an empirical value, assert that both are same targets, and carries out next step comparison; If degree of overlapping not in empirical value, confirms that both are not same targets, and abandon comparison;

(5), compare the degree of overlapping of both relevant positions in image with circularity in the step (3) and area; If degree of overlapping in an empirical value, assert that both are same targets, and carries out next step comparison; If degree of overlapping not in empirical value, confirms that both are not same targets, and abandon comparison;

(6), compare the degree of overlapping of both relevant positions in image with the peak in the step (2), minimum point coordinate and the wideest line segment width and position; If degree of overlapping in an empirical value, confirms that both are same targets, comparison finishes; If degree of overlapping not in empirical value, confirms that both are not same targets, the comparison EOP (end of program).