Real-time Infrared Image Processing And Tracking Algorithm Research Based On DSP

Infrared image processing and tracking of image processing is the hot topic, infrared image is different from visible light into the focal plane image, reflect the images of light and shade, therefore in the infrared image processing will often encounter the following problems: infrared images for computing the core requirements are higher processing speed, single DPS operation speed is not enough; Visual image processing to make target in the background image is not easy to distinguish; Image matching, a large amount of calculation and the problem of mismatch; Target detection, target and background segmentation threshold is difficult to determine; Target tracking, the tracking speed and effectiveness of the algorithm. In view of the above problem, this paper studies based on double DSP real-time infrared image processing and tracking, in order to realize the air precision of infrared target detection and recognition in the background.First, the article construct the image tracking system. In view of the main technical index requirements, use optical devices, infrared detectors, such as the DPS parts, using two DSP jointly solve the demand for speed, and expanding exhibition the SDRAM to store to solve large amount of data storage, the result of the operation to study actual image picture is fluent, can meet the technical index requirements.Second, the article study air targets image visual processing. Through the analysis of sky background and the target temperature range to get gray level histogram, the histogram to remove part of infrared detector detection range is too large, and variable threshold platform instead of histogram equalization and histogram method enhance the dynamic range of the target, making them easy to observe in the image and not lose the goal; The genetic algorithm is used in image matching, in order to avoid the problem of low efficiency of traditional genetic algorithm calculation, the research to the average normalized product correlation and genetic algorithm to solve real-time image matching error caused by non zero mean error. And search strategy of genetic algorithm in the image into multiple regions, each region has the existence of initial population in order to make sure to adapt to the overall situation.Finally, the research of air target detection and tracking. Information entropy is applied to detect, instead of one-dimensional and two-dimensional entropy, with the condition of maximizing the background and goal of entropy and selecting proper threshold for segmentation; Two-dimensional entropy calculations for big shortcoming, improve search strategy, searching for the target may after45of the probability that the point of maximum entropy, accurate search for entropy maximum points around, using recursion formula in the calculation of operation, shorten operation time. Research based on particle swarm filtering method of target tracking, using a posteriori probability instead of prior probability, sampling, using particle swarm optimization method to optimize the particle filter sample diversity and weight increase, increase the tracking accuracy.For algorithm research of the above test result, this paper built the system of software and hardware is other parts can satisfy the detection requirements, have good practical value and application prospect.