Research On Denoising And Enhancement Technology Of Night Vision Image For Apple Harvesting Robot

Apple is one of the most common fruits in people’s daily life, its growing area and yield are also increasing. Picking operation is an important part of the production of apple planting. At present, it has to be dependent on manual picking basically, which has labour-intensive, long time consumption and low efficiency. In recent years, the research of apple harvesting robot has received extensive attention. It also has made some progress, but the work efficiency is still to be improved. At the same time, with the increase of market demand, improving the working efficiency is needed to pick ripe apples timely. As a kind of highly automatic operation machinery, the harvesting robot can prolong its working time at night, and it has important practical significance to improve its working efficiency. The primary task of the robot to realize picking operation at night is the automatic target recognition of apple images. Because of the lack of light at night, the night vision image has many dark spaces, low contrast, shadows besides the low resolution. All this puts forward a new challenge to the harvesting robot. Based on the image processing technology, the key technologies of image denoising, enhancement and so on are studied in this paper.Firstly, it introduces the artificial light sources used by the acquisition of the night apple images. And the images will be analyzed from two aspects of subjective vision and histogram and its gradient change. Then it mainly introduces the wavelet transform denoising algorithm and sparse decomposition denoising algorithm after describing that how to evaluate image denoising effect. On this basis, using particle swam optimization(PSO) to improve sparse decomposition algorithm to denoise the night apple image is proposed. The improvement is that PSO algorithm is used to optimize the orthogonal matching tracking process, it makes search algorithm can quickly find and converge to the global minimun. At the same time, these different algorithms are compared through experimental data about their noise reduction effect. According to the deficiencies of the night, some research about image enhancement technology are done in this paper, while giving the advantages and disadvantages of the traditional image enhancement algorithms including histogram equalization, homomorphic filtering, bilateral filtering and so on.Meanwhile, Retinex algorithm based on guided filtering is proposed which can preserve the edge.Firstly, an apple night vision image of the RGB is converted into HSI color space. Then its intensity of the image is processed by the guided filter which has a function of edge-maintenance.This algorithm is able to accurately estimate the illumination of the image at the edge of high contrast. After that, a single scale Retinex algorithm is used for logarithmic transform to get thereflection image. Then, the Gamma corrects reflection component and illumination component.The two parts of the image are synthesized into a new image. Moreover, the contrast between the enhancement algorithm and other algorithms is given in this paper from two aspects of the subjective visual effect and objective quality performance parameters. Finally, after the processings of the image noise reduction and the image enhancement, the image is processed by Otsu threshold segmentation based on the R-G image, which will be compared with the threshold segmentation image without denoising and enhancement. After that, all the results are given. The experimental results show that the proposed algorithm in this paper can make the apple fruit is displayed completely in the dark region, and their edges are smooth and clear.