Research And Implementation Of Image Enhancement Of The Greenhouse Environment

In the greenhouse environment, the shadow image and low light image processing hasbeen one of the most difficulties in image processing.In order to solve these problems, it is notonly important in theory, also in practice it has a broad market demand.For example,when Thewater concentration in the air is too strong, the images of flowers or crops are not clear. Theseconditions have seriously affected the image visual quality for real-time monitoring, flowersor crop health and subsequent disease diagnosis.Therefore, the image enhancementtechnology of greenhouse environment has become the key technology in this field.Combined with the actual situation, this paper is mainly to solve the two problems above,andbased on the Linux embedded development platform, the final design of the embedded imageprocessing system based on QT4, which was used successfully in this paper about the imageenhancement algorithm, has achieved the expected results, In addition, the image processingsystem has the advantages of simple operation, good portability, and has a very broadapplication value.For fuzzy image problem, this paper which based on the combination of the basis of thetraditional algorithms, proposes a median filtering method based on wavelet transform, and itis a combination of time domain and frequency domain characteristics of the image analysis.Firstly, using median filtering for noise image, and then the filtered images are decomposedby wavelet transform, then the wavelet coefficient matrix generated;Using the medianfiltering principle to process new generation wavelet coefficient matrix; Finally, select thecorresponding wavelet threshold, denoising image, and then generate new image. For theproblem of image exposure, this paper studies the basic theory of Retinex and color constancy,and finally this paper proposed the Retinex enhancement algorithm based on local featuresand global features. In the first algorithm, the central idea is to calculate the relative value ofeach pixel, and the pixel gray level correction.then this paper improved the algorithm of linearstretching process, and achieved remarkable results. In the last algorithm, the central idea is toestimate the brightness of an image, then according to the corresponding formula, calculatedthe reflection part of the image, which is namely the enhanced image.Finally, based on the Linux system as embedded software platform, and combined withSEED-DVS6446experimental box, this paper designed the enhance system by QT4. Thesystem realizes the communication between ARM and DSP and the transplant in DVS6446byQT4.In addition, the system realizes the real-time image acquisition, real-time display andimage processing combining with the algorithm above.