Method and device for semi-automatically correcting target tracking frame, and electronic equipment

Abstract

The invention provides a method and a device for semi-automatically correcting a target tracking frame, and electronic equipment. The method comprises the following steps: receiving an operation instruction which is sent by an operator through launch control equipment and is used for determining a target position, amplifying an image corresponding to a target region, and displaying the amplified image on an original image in a picture-in-picture manner; when an operation instruction for correcting the target position sent by the operator through the launch control equipment is received, correcting the target region according to the operation instruction for correcting the target position; preprocessing the corrected image corresponding to the target region, segmenting targets in the target region, and finding out an optimal target; tracking the optimal target, if the operation instruction sent by an operator through the launch control equipment for adjusting the position of the target is received in a tracking process, initializing a tracker, and otherwise, performing normal tracking. According to the method and the device, the tracking precision of lock-turn tracking can be further improved, and the targets can be tracked more effectively and accurately.

Description

technical field [0001] The invention relates to the field of image tracking algorithms, in particular to a method, device and electronic equipment for semi-automatically correcting a target tracking frame. Background technique [0002] Object detection techniques are commonly used in object tracking systems. Target detection is generally divided into two stages, the first stage is the operator locks the target, and the second stage is the target tracking frame correction. One of the difficulties is to accurately detect the target, determine the position of the target tracking center and the size of the tracking frame. [0003] The image tracking system currently uses two types of input video images: TV and infrared, and the ground targets are generally vehicles, buildings and other targets. In the image tracking system, after the image is transmitted to the launch control receiver through optical fiber or wireless data link, the operator can identify the specific position ...

AI Technical Summary

Problems solved by technology

[0004] When the target is initially locked, due to the complex ground scene, the distance between the tracker and the target is long, and the target imaging pixels are few, rain, fog, snow, and sand and dust will seriously attenuate the energy radiated by the target, which greatly reduces the interception of the target by the tracker Distance; the ground background's reflection of sunlight and its own heat radiation make it difficult for the tracker to find the target in the background

Therefore, there may be a pixel deviation between the position of the tracking point and the position of the center point of the target when locking, and the initial deviation tends to accumulate continuously during the tracking process, and the error increases at the end of the tracking, which in turn affects the hit probability

Therefore, we need to solve problem 1: the problem of real-time detection, locking and recognition of weak and small targets with low signal-to-noise ratio in complex backgrounds

[0005] At the same time, as the distance between the tracker and the target approaches, due to the different approaching angles, the affine perspective relationship in the scene changes, and the three-dimensional relative positional relationship between the background around the target and the target is reflected in the relative positional relationship on the two-dimensional plane image. , there will be a series of complex situations such as the target blending into the background, the foreground occluding the target, and the shadow under the target's feet appearing, and the midpoint of the guide tracking deviates from the center of the target.

Therefore, we need to solve the second problem: human assistance is required to correct the target, otherwise it is easy to track the background and cause the target to be lost

Due to the complex ground background, the segmentation result is easily disturbed by the background information, resulting in segmentation errors, and errors are prone to occur in locking the center of the target

Therefore, the existing technology cannot guarantee the accuracy of initial target locking in the case of motion platform image shake

Since the initial locking error will continue to accumulate during the tracking process, it will affect the subsequent tracking process. Therefore, the existing technology cannot solve the initial locking error accumulation during the approach process, nor can it solve the target blending into the background caused by the change of the affine perspective relationship. , The target is occluded, etc. affect the tracking, resulting in the tracking frame offset problem

Images