| Real-time and rapid extraction of plant phenotype information has become a very critical step in the development of agricultural production,which can provide help for plant prediction and detection.Bananas are the fruits of the key development in Guangxi.Real-time measurement of the width and height of the pseudo-stem of bananas can provide effective help for the extraction of banana plant growth parameters and later yield evaluation.In this thesis,banana plants are taken as the research object,and the method of combining binocular stereo vision and cascade classifier is used to realize the rapid non-destructive measurement of the width and height of the pseudo-stem of banana plants.An image measurement method based on binocular camera is designed.The main research contents include:(1)The cascade classifier recognizes the pseudo-stem of bananas.According to different measurement purposes,two batches of experiments were carried out,using close distances(0.9 m,1.2 m,1.6 m,1.9 m,some pseudo-stem can be photographed),and long distances(1.8m,2.0 m,2.2 At m and 2.4m,the whole pseudo-stem can be photographed)Two methods were used to collect the binocular images of the pseudo-stem of bananas,and 100 samples were collected at each distance.Respectively established two models of banana pseudo-stem recognition model.First,disparity maps are obtained by dual-target targeting,stereo correction,and stereo matching,and they are randomly divided into a training set(170 samples)and a test set(230 samples).Then,a manual rectangular frame is used for the disparity map fake stem in the training set(the aspect ratio is 2: 1 in the short distance mode and the aspect ratio is 7: 1 in the long distance mode)to intercept positive and negative samples.Then,the cascade classifier is used to train the positive and negative samples to obtain the automatic recognition model of the pseudo-stem.The results show that this method can quickly identify the pseudo-stem of bananas,and the recognition rates of short distance and long distance are 94.8% and 96.1%,respectively.(2)A method for estimating the stem width of banana pseudo-stem is designed.This method focuses on the rectangular frame area identified by the cascade classifier,further accurately locates the boundary points of the pseudo-stem at different levels,and converts it to the camera coordinate system to obtain the horizontal width,and then corrects it by the inclination To get the final measured width of the pseudo-stem.The experimental results show that different measurement distances willaffect the detection accuracy.In the short-distance measurement mode,the measurement accuracy at the 1.2 m distance is the highest,and the R2 from the true value reaches 0.8538,and the relative error is only 2.0%.(3)A joint estimation method for banana stem height and stem width is designed.This method is mainly aimed at the situation that the complete pseudo-stem body has been obtained in the long-distance measurement mode.While measuring the stem width parameters in the same way in the previous step,the distance between the two points is also used to calculate the distance between the root of the false stem and the thinnest pseudo-stem under the lowest leaf branch.The results show that different measurement distances will affect the detection accuracy.The measurement accuracy of the stem height and stem width of the banana plant at a distance of 1.8 m is the highest,and the R2 of the stem height recognition model and the true value reaches 0.9717,and the relative error is only 3.1%;The stem width identification model and the true value R2 reached 0.9653,and the relative error was only 1.2%.In this paper,the method of binocular vision combined with cascade classifier is used to quickly identify the banana pseudo-stem,so as to achieve the purpose of obtaining the stem height and stem width of the banana pseudo-stem in real time,which improves the efficiency of measurement and provides technical support for the orchard information management. |
PDF Full Text Request