Simplify the interpretation of alert lists for clinical mastitis in automatic milking systems

Based on sensor measurements, an automatic milking system (AMS) generates mastitis alert lists indicating cows which are likely to have clinical mastitis (CM). Because of the general assumption of equal probabilities of developing CM for all cows, all alerts on the list have the same success rate. As a consequence, it is not possible to rank-order the alerts in terms of their likelihood of CM. In practice, the performance of a CM detection system is not only based on the sensitivity (SN) and specificity (SP) of the system, but is also influenced by the prior probability of a cow having CM. This study illustrates the idea of using cow-specific prior probabilities of CM, based on non-AMS information, to provide a rank-order on the alerts from an AMS. A tree-augmented naive Bayesian network was trained from available data to determine these cow-specific prior probabilities for CM. The graphical structure of the network and the probability tables for its variables in the network were based on data from 274 Dutch dairy herds that recorded each case of CM over an 18-month period. The final data set contained information on a total of 5363 CM cases derived from 28,137 lactations and 22,860 cows. The available prior cow information (parity, days in milk, season of the year, somatic cell count history and CM history) was included as variables in the network. By combining the cow-specific prior probabilities of CM with the SN and SP of the detection system of the AMS, the computed success rates can be used to discriminate between CM alerts. Our illustrations indicate that the success rate might range from 3 to 84%, while assuming an equal overall probability would result in a success rate of 21%. Using the computed success rates, the CM alerts on an alert list can be rank-ordered, thereby providing the dairy farmer information about which cows have the highest priority for visual inspection for CM.     

Evaluation of ultrasonic sensor for variable-rate spray applications

Automatic variable-rate sprayers require accurate measurement of canopy size. An estimate of canopy size is made by measuring the distance to the canopy at several elevations above the ground; an ultrasonic sensor was used to determine canopy distance in this study. It is sometimes necessary to conduct spray operations during harsh operating conditions. In this study ultrasonic sensors were subjected to simulated environmental and operating conditions to determine their durability and accuracy. Conditions tested included exposure to extended cold, outdoor temperatures, cross winds, temperature change, dust clouds, travel speeds and spray cloud effects. The root mean square (RMS) error in a series of measurements of the distance to a simulated plant canopy was used to test for significant difference among treatments. After exposure to outdoor cold conditions for 4 months, the RMS error in distance measured by the ultrasonic sensor increased from 3.31 to 3.55 cm, which was not statistically significant. Neither the presence of dust cloud nor the changes in cross-wind speeds over a range from 1.5-7.5 m/s had significant effects on the mean RMS errors. Varying sensor travel speed from 0.8 to 3.0 m/s had no significant influence on sensor detection distances. Increasing ambient temperature from 16.7 to 41.6 °C reduced the detection distance by 5.0 cm. The physical location of the spray nozzle with respect to the ultrasonic sensor had a significant effect on mean RMS errors. The mean RMS errors of sensor distance measurements ranged from 2.3 to 83.0 cm. The RMS errors could be reduced to acceptable values by proper controlling the sensor/spray nozzles spacing on a sprayer. In addition, multiple-synchronized sensors were tested for their measurement stability and accuracy (due to possible cross-talk errors) when mounted on a prototype sprayer. It was found that isolating the pathway of the ultrasonic wave of each sensor reduced detecting interference between sensors during multiple sensor operation. Test methods presented herein may be useful in the design of standardized testing protocols for field use distance sensors.     

Description of a detection model for oestrus and diseases in dairy cattle based on time series analysis combined with a Kalman filter

Sensor measurements can be used in dairy farming for the detection of oestrus and diseases. A new model has been developed to process the measured variables in a combined way. It is based on time series models for milk yield, milk temperature, electrical conductivity of quarter milk and the cow’s activity, and a probability distribution for the concentrates leftovers. The parameters of the time series models and the probabilities are fitted on-line for each cow after each milking by Kalman filters. This makes it possible to combine the variables and to generate cow-specific alerts. Global results on the detection of oestrus, mastitis and other diseases are given.     

基于多时相资源一号04星多光谱影像的土地利用分类——以海南西北部地区为例

【目的】基于3期2015年获取的资源一号04星(CBERS-04)多光谱遥感数据,探讨CBERS-04多光谱数据在热带地区土地利用分类中的应用潜力。【方法】结合光谱和物候信息,分别采用最大似然法和决策树分类方法对海南西北部地区土地利用现状进行分类研究。【结果】基于单景的最大似然法可获得相对理想的分类精度,总体分类精度为85.8%~88.8%,卡帕系数为0.80~0.84;同时使用3期影像作为输入,运用最大似然法和决策树分类方法,其分类精度均有明显提升,总体分类精度达91.61%~92.61%,卡帕系数为0.88~0.89,其中最大似然法略优于决策树分类算法。【结论】联合多期CBERS-04多光谱数据能够准确提取热带地区土地利用现状信息,具有广阔的应用前景。     

机器学习在近红外光谱法判别鲍鱼品种研究中的应用

为解决市场上鲍鱼产品缺乏科学分类方法的问题,利用近红外光谱分析技术结合机器学习方法对鲍鱼快速分类进行研究,使用MicroNIRTM1700便携式近红外光谱仪采集3种鲍鱼,即绿盘鲍(25只)、红壳鲍(31只)、皱纹盘鲍(35只)的光谱数据,采用CART算法建立鲍鱼分类决策树模型,以模型对测试集样本的预测准确率衡量决策树模型优劣,分裂策略为在每个节点处选择Gini不纯度最大的方式进行分裂,通过交叉验证控制决策树深度。结果表明,对训练集180条光谱建立模型,采用5折交叉验证,模型准确率为90.00%,对测试集93条光谱的预测准确率为90.32%。本研究方法可以很好地区分绿盘鲍、红壳鲍和皱纹盘鲍,满足鲍鱼现场快速分类的需求。     

Development of the Berry Impact Recording Device sensing system: Hardware design and calibration

Bruising caused by the impact damage occurs frequently during mechanical harvest process for highbush blueberries. The overall goal of this study was to develop a miniature and low-cost sensor prototype to quantitatively measure the impact forces endured by blueberries during the mechanical harvest process, which could be used to reduce blueberry bruising through improved harvester design. The sensing system developed in this study had three essential components: the sensor, the interface box, and the computer software program. The round circuit board of sensor is less than one inch (19.4 mm), including three accelerometers with ±500 g sensing range in each orthogonal axis, one eight-bit microcontroller, one 128 KB memory chip, and other electronic components with low power consumption. The sensor board and rechargeable battery were cast into a one inch (25.4 mm) sphere using silicone rubber. The interface box serves as the intermediate communication platform to connect the sensor and the computer. The PC-software retrieves data from the sensor via the I²C communication and downloads data to a computer for further analysis via the RS232 communication. The sensor was calibrated using a centrifuge. The accuracy of the sensor output was 0.53% (2.60 g maximum deviation) and −0.33% (−1.26 g maximum deviation), with precision error of 0.63% (3.21 g) in the output span. This miniature and low-cost sensor prototype provides the opportunity to understand how the berry (or other small fruits) interacts with different machine parts within the harvester and to identify critical control points that cause the most mechanical impacts, which was not achievable in the past.     

Assessment of forage mass from grassland swards by height measurement using an ultrasonic sensor

The non-destructive assessment of forage mass in legume-grass mixtures as a tool for yield mapping in precision farming applications has been investigated in two field experiments. An ultrasonic sensor was used to determine sward heights. Forage mass-height relationships were evaluated by carrying out static measurements on binary legume-grass mixtures of white clover (Trifolium repens L.), red clover (Trifolium pratense L.), and lucerne (Medicago sativa L.) with perennial rye grass (Lolium perenne L.) across a wide range of sward heights (5.0-104.2 cm) and forage mass (0.15-11.25 t ha⁻¹). Mobile measurements, hereafter referred to as “on-the-go” were conducted by mounting the ultrasonic sensor in combination with a high-precision Differential Global Positioning System (DGPS) on a vehicle. Data were recorded along experimental plots consisting of perennial rye grass and grass-clover mixtures similar to the mixtures that were used for the static experiment. The static experiment revealed a relationship between ultrasonic sward height and forage mass explaining 74.8% of the variance with a standard error (SE) of 1.05 t ha⁻¹ in a common dataset. The type of legume species, weed proportion, and growth period had a significant impact on the above mentioned relationship. Legume-specific regression functions had higher R²-values of up to 0.855 (white clover mixture). Datasets including legume-specific mixtures and pure swards of both components reached comparable R² values between 0.799 and 0.818 but exhibited higher SE values. The abundance of weeds resulted in increased ultrasonic sward heights for the same levels of forage mass. On-the-go measurements across experimental field plots yielded a sward height range of 1.4-70.4 cm. Abrupt forage mass changes at the transition from treatment plots to cut interspaces resulted in a significant deviation from stubble height within a distance of 50 cm to plot borders. When legume-specific equations derived from static measurements were applied to sward heights, forage mass was overestimated by 21.4% on average. Mean residuals from predicted forage mass ranged between 0.893 (pure grass) and 1.672 (red clover mixture) and increased significantly if the point sampling distance along the track was increased to more than 0.82 m on average across all plots. The prediction accuracy of forage mass from ultrasonic height measurements is promising; however, further modifications to the technique are necessary. One such improvement can be the use of spectral reflectance signatures in combination with the ultrasonic sensor.     

A low-cost microcontroller-based system to monitor crop temperature and water status

A prototype system was developed and constructed for automating the measurement and recording of canopy-, soil-, and air temperature, and soil moisture status in cropped fields. The system consists of a microcontroller-based circuit with solid-state components for handling clock/calendar, sensor power, and data storage and retrieval functions. Sensors, including an analog soil moisture sensor, analog and digital temperature sensors, and a digital infrared thermometer, are widely available and inexpensive. The circuit board and sensor assemblies require approximately 4 h to construct and test, and material costs totaled approximately US$84. Systems were built and tested during the 2009 growing season in a corn field to evaluate performance and suitability under local conditions. The sensors performed according to manufacturers’ specifications, with accuracies of ±0.4 °C, ±1.4 °C, and ±0.3 °C for air-, soil-, and canopy-temperature measurements, respectively. Soil moisture sensors were calibrated and provided measurements within ±2 kPa of the manufacturer's values. Reliability of data collection and storage averaged 91%, with most bad or missing data occurring during periods of inclement weather and electrostatic interference.     

绵阳市奶牛乳房炎的调查与分析

采用临床调查、BMT法和苛性钠法对绵阳市6个不同规模奶牛场的乳房炎进行了检测。结果表明，临床型乳房炎的发病率为2％～11．67％（平均发病率为6．08％），隐性乳房炎的感染率为41．67％～86．40％（平均感染率为55．00％）。用SPSS统计软件对奶牛乳房炎各乳区之间的相关性及BMT和苛性钠两种检测方法的差异性进行了方差分析。结果表明，奶牛乳房炎各乳区之间的感染率差异不显著（P〉0．05），乳房炎的发生与乳区之间无明显相关性；BMT和NaOH两种方法对乳房炎检测的结果之间差异不显著（P〉0．05），表明这两种方法都可用于隐性乳房炎的检测。     

Study on Soil Salinization Information in Arid Region Using Remote Sensing Technique

Extracting information about saline soils from remote sensing data is useful, particularly given the environmental significance and changing nature of these areas in arid environments. One interesting ease study to consider is the delta oasis of the Weigan and Kuqa rivers, China, which was studied using a Landsat Enhanced Thematic Mapper Plus (ETM+) image collected in August 2001. In recent years, decision tree classifiers have been successfully used for land cover classification from remote sensing data. Principal component analysis (PCA) is a popular data reduction technique used to help build a decision tree; it reduces complexity and can help the classification precision of a decision tree to be improved. A decision tree approach was used to determine the key variables to be used for classification and ultimately extract salinized soil from other cover and soil types within the study area. According to the research, the third principal component (PC3) is an effective variable in the decision tree classification for salinized soil information extraction. The research demonstrated that the PC3 was the best band to identify areas of severely salinized soil; the blue spectral band from the ETM+ sensor (TM1) was the best band to identify salinized soil with the salt-tolerant vegetation of tamarisk (Tamarix chinensis Lour); and areas comprising mixed water bodies and vegetation can be identified using the spectral indices MNDWI (modified normalized difference water index) and NDVI (normalized difference vegetation index). Based upon this analysis, a decision tree classifier was applied to classify landeover types with different levels of soil saline. The results were checked using a statistical accuracy assessment. The overall accuracy of the classification was 94.80%,which suggested that the decision tree model is a simple and effective method with relatively high precision.     

张家口地区奶牛乳房炎调查

采用SMT检测法,对张家口地区108头泌乳奶牛进行乳房炎检测.结果显示,该地区奶牛隐性乳房炎头阳性率为47.33%,临床型乳房炎头阳性率为5.56%,乳区阳性率为40.99%(158/422).     

聚类、粗糙集与决策树的组合算法在地力评价中的应用

 【目的】地力评价方法大多数有一定的主观性,较少考虑土壤各属性间的依赖关系。论文旨在采用数据挖掘方法,寻求地力等级划分的新方法。【方法】结合农安县耕地调查数据,应用K-means聚类方法、Johnson粗糙集属性约简算法与C4.5决策树算法相结合的优化算法评价地力等级。【结果】使用K-means聚类方法,得到最佳学习样本数;使用粗糙集属性约简和决策树相结合的方法,去掉了冗余属性7个,决策树模型共有节点317个,其中叶节点个数为159个,生成规则159条,模型准确率为82.08%。与未聚类和未约简的方法相比,决策树结点个数减少41.62%。【结论】使用该组合算法,在保证模型准确率的同时,降低了算法的时间和空间复杂性,提高了挖掘效率。     

Development of a low-cost attitude sensor for agricultural vehicles

The objective of this research was to develop a low-cost attitude sensor for agricultural vehicles. The attitude sensor was composed of three vibratory gyroscopes and two inclinometers. A sensor fusion algorithm was developed to estimate tilt angles (roll and pitch) by least-squares method. In the algorithm, the drift error of the gyroscopes was estimated using the inclinometers. In addition to tilt angles, the attitude sensor also estimated the absolute heading angle and position with inclination error correction by integrating a GPS. Tests were conducted on a flat field, a sloping ground and a bumpy road. Results showed that the attitude sensor was able to estimate the roll angle with the maximum root mean square error of 0.43°, the pitch angle with 0.61° and the heading angle with 0.64°. Moreover, the attitude sensor dramatically improved the positioning accuracy from 25.9 cm to 3.0 cm in the sloping ground test and from 8.4 cm to 3.7 cm in the bumpy road test. The proposed technology used in the attitude sensor will help to develop advanced agricultural applications.     

Development of the Berry Impact Recording Device sensing system: Software

This paper reports a complete impact data acquisition, processing, and analyzing software system that applies on the hardware platform of the Berry Impact Recording Device (BIRD). The software has three major sections that correspond to the hardware: The BIRD sensor program, the interface box program, and the computer software i-BIRD. The sensor program samples acceleration data from three axes and records them as single impacts with a maximum sampling rate of 3.0 kHz. Users can configure the sensor via the i-BIRD computer software, with different options of sampling frequencies (682-3050 Hz) and thresholds (0-205 g, where g is the gravitational acceleration). The data recorded can be downloaded, processed and graphically displayed on the computer. A real time clock was created using the interrupt service routine provided by the microcontroller. The accuracy of the sensor’s clock was calibrated with an error of 0.073%, which was adequate to record impact data in this application. The shape of impact curves recorded by the BIRD sensor at three sampling frequencies (682, 998, and 1480 Hz) matched well with the curves recorded by a high frequency (10 kHz) data logger with the maximum root mean squared error of 4.4 g. The velocity change had a relative error less than 5%. With confirmation of all those performances, the software system enabled the BIRD to be a useful tool to collect impact data during small fruit (such as blueberry) mechanical harvest.     

基于TWDTW的时间序列GF-1 WFV农作物分类

【目的】焉耆盆地是新疆重要的特色农产品生产基地,农作物种植结构较为复杂。利用时间序列的遥感数据对研究区内的农作物进行分类识别,获得不同农作物的空间分布、种植面积等信息,为政府部门制定粮食政策、经济计划提供重要依据。同时探讨时间加权的动态时间弯曲(time weighted dynamic time warping,TWDTW)方法在农作物分类识别中的适用性以及高分一号(GF-1)WFV在农业领域的应用潜力。【方法】以新疆焉耆盆地为研究区域,利用2018年作物生长季的GF-1 WFV时间序列数据集计算归一化植被指数(NDVI),基于TWDTW方法开展农作物分类识别研究。分别采集不同作物的样本点,形成各作物NDVI的标准序列。利用TWDTW相似性匹配算法计算每个待分类像元与不同作物标准序列间的相似度距离,距离值越小则相似性越高,通过对比确定像元的农作物类型,得到最终的分类结果,同时根据时间序列NDVI曲线建立决策树(decision trees,DTs)分类规则,人工设置分类阈值得到分类结果,并与TWDTW方法的结果进行对比分析。【结果】2种方法的分类结果较为一致,辣椒的种植范围最广,小麦主要分布在焉耆盆地北部和西部的农二师二十一团,番茄和甜菜的种植分布较为零星。在种植面积统计中,辣椒的种植面积最大,其后依次为番茄、小麦和甜菜。利用野外样本点对决策树和TWDTW两种方法的分类结果进行精度验证,总体精度分别为89.58%和90.97%,kappa系数为0.804和0.830,TWDTW方法的分类精度相比于决策树法略有提高。【结论】相比于决策树分类方法,TWDTW方法的分类精度略有提高的同时,分类结果客观可靠,而且算法不受地域因素限制,具有较强的灵活性和适用性。基于密集时相的GF-1 WFV数据集,采用TWDTW算法对农作物进行分类,得到较好的分类结果,能够满足农业部门的管理决策需求,该方法在农业领域具有较大的应用和推广价值。     

Crop Identification Based on TWDTW Method and Time Series GF-1 WFV

【Objective】 Yanqi Basin is an important production base of characteristic agricultural products in Xinjiang, and the planting structure of crops is complicated. In this study, the time series remote sensing data were used to classify and identify crops in the study area, so as to obtain the spatial distribution of different crops and their planting areas, which were the important basis for government sectors to formulate grain policies and economic plans. At the same time, the applicability of time-weighted dynamic time warping (TWDTW ) method in crop classification and the application potential of GF-1 WFV in agriculture were also discussed.【Method】 The normalized vegetation index (NDVI), calculated from the 2018 time series GF-1 WFV data set in Yanqi Basin, Xinjiang, was used to study the crops recognition based on TWDTW method. Sample points of different crops were collected to form standard sequence of NDVI for each crop. The TWDTW similarity matching algorithm was used to calculate the similarity distance between each pixel to be classified and the standard sequence of different crops. The smaller the distance was, the higher the similarity was. The similarity was used to determine the crop type of the pixel, and the final classification result was obtained. At the same time, the classification rules of decision tree were established according to the NDVI curve of time series, and the classification result was obtained by manually setting the classification threshold, and compared with that of the TWDTW method. 【Result】 The classification results of the two methods were very consistent. Peppers were the most widely planted and the wheat was mainly distributed in the northern part of the Yanqi Basin and the 21st Division of the Second Agricultural Division. The distributions of tomato and sugar beet were relatively sporadic. Among the results of planting area, pepper had the largest planting area, followed by tomato, wheat and sugar beet. The accuracy of the classification results of the TWDTW and decision tree methods was verified by the field sample points: The overall accuracy of them were 89.58% and 90.97%, respectively, and the kappa index of them were 0.804 and 0.830, respectively. The classification accuracy of the TWDTW method was slightly higher than that of the decision tree method. 【Conclusion】 Compared with the decision tree classification method, the classification accuracy of the TWDTW method was slightly improved, the classification result was more objective and reliable. The algorithm of TWDTW method was not limited by geographical factors and had strong flexibility and applicability. The experimental results showed that using TWDTW algorithm to identify crops based on the GF-1 WFV data set of dense temporal phase could get better classification results, and it had great application and popularization value in agricultural field.     

茶树油对LPS诱导的奶牛乳腺炎的作用及其机制

【目的】奶牛乳腺炎一直是奶牛养殖业和奶制品行业最大的挑战之一,制约着奶业的健康发展。有效进行奶牛乳腺炎的防治,可以为奶牛的健康、生产优质乳制品提供良好保障。探究茶树油对LPS诱导的奶牛乳腺炎作用效果,探索茶树油替代抗生素治疗奶牛乳腺炎的可行性,为茶树油治疗奶牛乳腺炎提供参考。【方法】在牛乳腺上皮细胞的培养中分别添加50、100、200、500、1 000μg·mL^-1的LPS进行相关指标的检测。通过CCK-8法检测细胞增殖活性、流式细胞仪检测细胞凋亡、实时荧光定量检测细胞因子表达量以及ELISA检测相关蛋白的表达量等方法检测乳腺上皮细胞的相关指标。研究构建了LPS诱导的奶牛乳腺炎模型。在200μg·mL^-1 LPS诱导12h的奶牛乳腺炎细胞模型中分别添加0.0002%、0.0004%、0.0006%、0.0008%、0.001%、0.002%、0.004%、0.006%、0.008%、0.01%的茶树油进行相关指标的检测。【结果】CCK-8法检测细胞增殖活性,结果显示100μg·mL^-1的LPS攻毒情况下,细胞已开始出现...     

Using a vision sensor system for performance testing of satellite-based tractor auto-guidance

A vision sensing system for the measurement of auto-guidance pass-to-pass and long-term errors was implemented to test the steering performance of tractors equipped with auto-guidance systems. The developed test system consisted of an optical machine vision sensor rigidly mounted on the rear of the tested tractor. The center of the drawbar hitch pin point was used as the reference from which to measure the deviation of the tractor's actual travel path from its desired path. The system was built and calibrated to a measurement accuracy of better than 2 mm. To evaluate the sensor, two auto-guidance systems equipped with RTK-level GNSS receivers were tested and the results for different travel speeds compared. Pass-to-pass and long-term errors were calculated using the relative positions of a reference at a collocated point when the tractor was operated in opposite directions within 15 min and more than 1 h apart, respectively. In addition to variations in speed, two different auto-guidance steering stabilization distances allowed for comparison of two different definitions of steady-state operation of the system. For the analysis, non-parametric cumulative distributions were generated to determine error values that corresponded to 95% of the cumulative distribution. Both auto-guidance systems provided 95% cumulative error estimates comparable to 51 mm (2 in.) claims and even smaller during Test A. Higher travel speeds (especially 5.0 m/s) significantly increased measured auto-guidance error, but no significant difference was observed between pass-to-pass and long-term error estimates. The vision sensor testing system could be used as a means to implement the auto-guidance test standard under development by the International Standard Organization (ISO). Third-party evaluation of auto-guidance performance will increase consumer awareness of the potential performance of products provided by a variety of vendors.     

Near-infrared spectroscopic sensing system for on-line milk quality assessment in a milking robot

A near-infrared (NIR) spectroscopic sensing system was constructed on an experimental basis. This system enabled NIR spectra of raw milk to be obtained in an automatic milking system (milking robot system) over a wavelength range of 600–1050 nm. Calibration models for determining three major milk constituents (fat, protein and lactose), somatic cell count (SCC) and milk urea nitrogen (MUN) of unhomogenized milk were developed, and the precision and accuracy of the models were validated. The coefficient of determination (r²) and standard error of prediction (SEP) of the validation set for fat were 0.95 and 0.25%, respectively. The values of r² and SEP for lactose were 0.83 and 0.26%, those for protein were 0.72 and 0.15%, those for SCC were 0.68 and 0.28 log SCC/mL, and those for MUN were 0.53 and 1.50 mg/dL, respectively. These results indicate that the NIR spectroscopic system can be used to assess milk quality in real-time in an automatic milking system. The system can provide dairy farmers with information on milk quality and physiological condition of an individual cow and, therefore, give them feedback control for optimizing dairy farm management. By using the system, dairy farmers will be able to produce high-quality milk and precision dairy farming will be realized.     

基于CART决策树的自然保护区植被类型分布研究

目的针对保护区监测需求，充分发挥GF-1 WFV影像的宽幅特点和面向对象、机器学习算法在遥感影像分类中的优势，提高保护区植被类型遥感监测的精度，为保护区管理决策提供依据。方法以甘肃省白水江国家级自然保护区为研究区，主要数据源包括GF-1 WFV多光谱数据、Landsat-8 OLI遥感数据、DEM数据、野外调查数据等。首先，对GF-1 WFV数据进行多尺度分割，将研究区划分为诸多区域性的分割对象；然后，以分割对象为基本单元，研究光谱特征、几何特征、纹理特征不同组合情况下，基于CART决策树分类的结果；最后，利用训练样本建立基于TTA的精度检验，并基于混淆矩阵对分类结果进行分析。结果在多尺度分割过程中，形状因子、紧致度分别设置为0.2和0.5时地物边界显示较好；当形状因子和紧致度固定时，研究区最佳分割尺度为40。精度检验结果表明，基于CART决策树的保护区植被类型分类结果整体精度均在83%以上，Kappa系数在0.80以上，优于最邻近分类法和支持向量机分类算法，其中基于光谱特征、几何特征、纹理特征的CART决策树分类结果精度最高，总体精度为85.18%，Kappa系数为0.832 2，优于光谱特征分类、光谱特征结合几何特征分类的方法。结论基于CART决策树算法和面向对象方法的GF-1遥感影像分类方法适用于保护区植被类型分布研究，可有效辅助保护区监测工作。