Vision-based measuring method for individual cow feed intake using depth images and a Siamese network

Feed intake is an important indicator to reflect the production performance and disease risk of dairy cows,which can also evaluate the utilization rate of pasture feed.To achieve an automatic and non-contact measurement of feed intake,this paper proposes a method for measuring the feed intake of cows based on computer vision technology with a Siamese network and depth images.An automated data acquisition system was first designed to collect depth images of feed piles and constructed a dataset with 24150 samples.A deep learning model based on the Siamese network was then constructed to implement non-contact measurement of feed intake for dairy cows by training with collected data.The experimental results show that the mean absolute error(MAE)and the root mean square error(RMSE)of this method are 0.100 kg and 0.128 kg in the range of 0-8.2 kg respectively,which outperformed existing works.This work provides a new idea and technology for the intelligent measuring of dairy cow feed intake.

Nitrogen partitioning and microbial protein synthesis in lactating dairy cows with different phenotypic residual feed intake

Background: Residual feed intake(RFI) is an inheritable measure of feed efficiency that is independent on level of production. However, physiological and metabolic mechanisms underlying divergent RFI are not fully elucidated.This study was conducted to investigate dietary nitrogen(N) partitioning and microbial protein synthesis in lactating dairy cows divergent in phenotypic RFI.Results: Thirty Holstein dairy cows(milk yield = 35.3 ± 4.71 kg/d;milk protein yield = 1.18 ± 0.13 kg/d;mean ± standard deviation) were selected for the experiment to derive RFI. After the RFI measurement period of 50 d, the 10 lowest RFI cows and 8 highest RFI cows were selected. The low RFI cows had lower dry matter intake(DMI, P < 0.05) than the high RFI cows, but they produced similar energy-corrected milk. The ratios of milk to DMI(1.41 vs. 1.24, P < 0.01) and energy-corrected milk to DMI(1.48 vs. 1.36, P < 0.01) were greater in low RFI cows than those in the high RFI cows. The low RFI cows had lower milk urea nitrogen than that in the high RFI cows(P = 0.05). Apparent digestibility of nutrients did not differ between two groups(P > 0.10). Compared with high RFI animals, the low RFI cows had a lower retention of N(5.72 vs. 51.4 g/d, P < 0.05) and a higher partition of feed N to milk N(29.7% vs. 26.5%, P < 0.05).Conclusions: The results suggest that differences in N partition, synthesis of microbial protein, and utilization of metabolizable protein could be part of the mechanisms associated with variance in the RFI.

Effects of Supplementary Feed with Different Apple Pomace-to-Alfalfa Meal Ratios on Milk Performance of Dairy Cows

[Objective] To study the effects of apple pomace on milk performance of dairy cows. ~Method] A total of 48 Holstein dairy cows at the same parity and with close lactation stage and similar daily milk yield were randomly assigned to six groups. The supplementary feed was composed of apple pomace, alfalfa meal, com and premix, and it was processed into pellets. In the groups I - V, the percentage of apple pomace was 17.5%, 35.0%, 52.5%, 70.0% and 0%, respectively. Each cow was fed with supplementary feed （3 kg/d）. [ Result] The milk yield in the group II and group V was increased significantly （ P 〈 0.05）, and the fat content in milk of the group III and V was significantly higher than that of the con- trol group （P 〈 0.05）. [ Conclusion] The supplementary apple pomace can improve the milk performance of dairy cows effectively.

Evaluation of Agolin^®, an Essential Oil Blend, as a Feed Additive for High Producing Cows

There is a growing demand for feed additives that can not only reduce dairy enteric methane emissions but also increase milk production and feed efficiency. Just one product is currently commercially available which accomplishes both of these goals. The purpose of this study was to confirm the performance benefits of the product (Agolin?, Agolin SA, Biere, Switzerland) in high producing mid-lactation dairy cows under United States feeding conditions. Four matched pens of approximately 150 mid-lactation cows/pen and averaging over 50 kg of milk/cow/day were enrolled in a side-by-side study. All pens received a common total mixed diet ad libitum, and the essential oil blend was administered via a concentrated farm pack to provide 1 g/cow/day to cows in the 2 test pens. Milk weights were determined, and samples were collected for compositional analysis over the last 2 days of the pretrial (May 11 and 12, 2020) and end of the trial (July 18 and 19, 2020) periods. Dry matter intake was measured by pen daily for the last 10 days of each feeding period. Milk fat and milk protein yields were greater (P < 0.05) for cows receiving added Agolin. There was a tendency (P = 0.06) for energy corrected milk/dry matter intake to be greater for cows receiving the Agolin (1.88) relative to the control diet (1.76). The trial showed that Agolin assisted in improving production parameters of economic importance to dairy producers.

Productive Response and Reproductive Performance of Dairy Cows Subjected to Different Feeding Systems

Fifty Holstein cows (2.6 ± 1.0 lactations) in early lactation (52.6 ± 16.4 days in milk) calved in autumn and producing 34.3 (±4.4) kg milk per cow·day-1 were randomly assigned for 27 weeks to one of two treatments (feeding systems) in 10 groups (pens) of 5 animals each (5 groups/treatment). The confined treatment was a total mixed ration (TMR_100) whereas the supplemented grazing system was defined by a partially mixed ration (PMR_75) with 75% TMR and 25% oat pasture (Avena sativa L.) during the autum-winter (Period I) and alfalfa pasture (Medicago sativa L.) plus concentrate (7.0 kg per cow·day-1) (P + C) during the next spring (Period II). Milk production was daily and individually recorded throughout the trial while the chemical composition of milk was measured on individual samples every 2 weeks. Live weight (LW) and body condition score (BCS) using a scale of 1 to 5 were recorded every 3 weeks after the morning milking. Simultaneously, blood samples were taken for determinations of glucose, urea, non-esterified fatty acids (NEFA), insulin, somatotrophin (GH) and somatomedin C (IGF-I). Total DM intake was measured using the difference method four times per repetition in the months of July, August, September and October. The experimental herd was inseminated at fixed time at the start of the service (May-July) and on the return with estrus detection. The diagnosis of pregnancy was made by transrectal ultrasonography at 30 and 60 days after the start of the service. The productive data were analyzed according to a model with repeated observations in time adjusted by covariate with 2 periods and using the animal as the experimental unit. Data of DM intake were analyzed using ANOVA for 2 treatments with 5 repetitions. Cows from the TMR_100 group produced more milk (kg&#183;cow-1&#183;day-1) during Period I (33.7 vs. 32.3) and Period II (28.8 vs. 27.8) (P &#183;cow-1&#183;day-1) in Periods I (1.33 vs. 1.24) and II (1.10 vs. 0.92) and for milk protein (1.19 vs. 1.13 and 0.99 vs. 0.93). The lower energy intake of grazing cows (45.0 vs. 43.1 and 40.9 vs. 38.9 Mcal per?cow·day-1 for TMR_100 vs. PMR_75 and TMR_100 vs. P + C) coupled to the extra energy cost associated with grazing activity would explain the difference in milk production. Milk protein content (%) resulted higher (P kg&#183;cow-1&#183;day-1) in the cows of the TMR_100 group in both periods (0.69 vs. 0.34 and 0.49 vs. -0.22) was consistent with the increase in the plasmatic levels of glucose and IGF-I and the reduction in the circulating levels of GH, but not with the lack of increases in plasma insulin concentration or decreases in circulating levels of NEFA, parameters that were not affected by treatments. Despite the positive effects of TMR_100 on parameters linked to energy balance, no significant differences were detected in any of the reproductive parameters evaluated and the final rate of pregnancy was 80.0 and 91.3% for the confined and grazing system with supplementation respectively. Free milk liters over feeding costs were higher in the grazing system with supplementation compared to the confined system (18.5 vs. 15.5 l, respectively). The results indicate that the grazing system with supplementation can be economically competitive compared to the confined system of milk production.

Evaluation of Feeding Dairy Cow with Whole Wheat Hay and Alfalfa

The preliminary results of this laboratory (unpublished) indicate that the nutritional value of whole wheat hay (wheat hay, for short) is very high, and the milk stage is the best period of wheat hay harvest. In this study, we investigated the feeding effect and economic benefits using wheat hay instead of alfalfa diet for dairy cows under the condition of the same energy and crude protein levels. Three types of diets were used: alfalfa diet, wheat hay diet and alfalfa + wheat hay diet. The results showed that the dry matter digestibility of alfalfa diet and alfalfa + wheat hay diet did not differ significantly (P > 0.1), but was significantly higher than that of wheat hay diet (P < 0.05). The wheat hay diet could produce more propionic acid and ammonia nitrogen (P < 0.05) in the rumen, and reduce the ratio of acetic/propionic and nitrogen utilization. There was no significant difference in milk production among the three diets (P > 0.1). There was no significant difference in milk somatic cell count and body condition score among the three groups (P > 0.1). The wheat hay diet could significantly increase milk protein and lactose (P contents of interleukin-6 in cows fed alfalfa diet and alfalfa + wheat diet were significantly higher than that in cows fed wheat hay diet (P < 0.05). There was no significant difference in interleukinm-6 between in cows fed alfalfa diet and alfalfa + wheat diet (P > 0.05). The use of wheat hay to replace imported alfalfa in whole or in part could save feed costs. Full substitution of alfalfa with wheat hay could have a daily economic benefit of 13.74 yuan.

Influence of Mango Feed-Based Diets on the Production and Economic Profitability of Montbeliarde Cows’ Milk at the End of Lactation in Burkina Faso

The study evaluated the effects of a diet incorporating a mango-based feed on the technical-economic performance of milk production of Montbeliard cows at the end of lactation in a dairy farm in Bobo-Dioulasso, Burkina Faso. To do so, 14 cows were subjected to 3 diets, including an experimental diet incorporating mango (ED) in comparison to a balanced control diet (BCD) and to the feeding practice of the farmer (FCD) during 60 days. Results show that the average amount of milk milked per day from ED (10.93 ± 0.17 L) was significantly higher (p < 0.05) than BCD (8.80 ± 0.19 L) and RTA (5.39 ± 0.14 L). The conversion index of FPD (4.24 ± 0.14) was significantly higher (p < 0.05) than that of BCD (1.75 ± 0.19) and ED (1.26 ± 0.16). Average feed costs per liter of milk (FC/L) were significantly different (p (125.95 ± 4.56 FCFA), BCD (83.67 ± 6.36 FCFA) and ED (60.06 ± 5.49 FCFA). The average gross profit margins (GPM) were significantly higher and decreasing (p ,953 ± 5031 FCFA), BCD (167,157 ± 5831 FCFA) and FCD (74,188 ± 4185 FCFA). Mango feed can be used in the diet of dairy cows. Its use contributes to the reduction of production costs and the generation of substantial profits. The valorization of mango residues is an alternative for a greater availability of feed at lower cost for the animals. It could contribute to a better management of processing residues that pollute the environment and to the reduction of fruit flies that affect orchard yields through reinfestation. The increased use of mango in dairy cattle feed should be encouraged.

Evaluation of Agolin® Ruminant, an Essential Oil Blend, as a Feed Additive for Cows at Two Levels of Production

Agolin® Ruminant (Agolin) is a commercially available blend of essential oils which has been demonstrated to reduce greenhouse gas emissions in dairy cows and improve energy corrected milk and feed efficiency. Further trials are needed with large numbers of cows to confirm the magnitude of effect of this feed additive on milk production under differing feeding conditions and stages of lactation. Information that can be quantified from newer on-farm systems of measurement is likewise needed. This study was conducted to determine the effects of this additive on high producing (48 kg of milk/day average) just past peak lactation and medium producing dairy cows (43 kg of milk/day average) further along in lactation under commercial conditions that would typically occur in the Pacific Northwest USA. Four pens of approximately 400 Holstein cows/pen (two high producing pens and two medium producing pens) were available for this side-by side study. Performance results were determined using data derived from a subscription standardized monitoring system (High Desert Dairy Laboratory, Inc, Nampa, ID, USA) that provided results for milk yield, fat percentage and protein percentage. As well daily in-stall electronic monitoring of milk only was available. All pens received a common total mixed ration typical of rations fed in the Pacific Northwest, USA. The Agolin feed additive was dispensed through the bulk mineral supply with control cows receiving the normal minerals and the treatment cows receiving the normal minerals plus Agolin. The trial began on August 24, 2021, with treatment cows provided with 1 g/head/day for the duration of the 8-week long study. Dry matter intakes were determined daily for the week before the trial was conducted and the last week of the trial. Treatment results were compared using a general linear model that considered pretrial milk, fat yield and protein yield, days in milk and lactation number. There were 678 high producing cows available for the duration of the study. Based on the standardized testing results, milk yield was greater (1.12 kg/cow/day;P < 0.05) for the cows in the treatment group. There was a tendency for fat corrected milk (FCM) and energy corrected milk (ECM) to be greater for the cows receiving the Agolin feed additive (0.96 and 0.86 kg/cow/day for FCM and ECM respectively, P < 0.10). There were 646 medium producing cows that participated in the trial. There was no treatment effect upon milk production (P = 0.27). There was a significant (P < 0.05) increase in milk fat yield and no change (P = 0.33) in protein yield for the treatment cows in this test group, resulting in greater FCM and ECM (1.12 and 0.95 kg/cow/day;P < 0.05). Daily milk monitoring resulted in a reduced decline in milk yield from the week before to the final week of the trial revealing greater persistency of milk for Agolin-fed cows in both high and medium production pens (1.58 and 2.13 kg/cow/day;P < 0.01). Likewise, overall feed efficiency was improved by 5.3% (P < 0.05) with the test product.

Agolin Ruminant®, an Essential Oil Blend, Increases Energy-Corrected Milk and Feed Efficiency in a High Component Dairy Herd

Agolin® Ruminant (Agolin) is a blend of essential oils developed to improve feed efficiency while at the same time reducing rumen enteric methane production. Studies have shown that the product improves lactational performance, but the range of results has been varied. This experiment evaluated the effects of the feed additive Agolinon milk production in a high-component (fat and protein) dairy herd when provided for an extended time. The experiment was conducted at a large commercial dairy in the Pacific Northwest, USA. Eight pens of cows (350 to 500 cows/pen) were blocked by production, and pens within blocks were randomly assigned to treatment. All pens received a common total mixed ration. Treatment consisted of providing 1 g/cow/day of Agolin to the test group. A 4-week adaptation period was followed by an 11-week study period. Milk production and milk composition were determined by cow by averaging daily performance for the one-week period before the start of the trial (covariate period) and the last week of each of the two test periods. Feed intakes were determined by pen at the same time. Milk yield was 1.11 kg/cow/day greater (P < 0.001) at the midpoint and 1.48 kg/cow/day greater (P < 0.001) at the end of the test period for cows receiving the feed additive. Similarly, protein yields were 0.03 (P < 0.001) and 0.07 kg/cow/day greater (P < 0.001) at the midpoint and end of the feeding period. Milk fat yield was not different (P = 0.854) between treatment groups at the midpoint of the trial but increased with treatment (P = 0.002) by the end of the trial. Energy-corrected milk/feed ratio, based on pen parameters, was significantly altered by treatment at the final test period (1.48 vs. 1.64 kg/kg for control and test treatments, respectively).

热应激对奶牛生产和健康的影响及其减缓措施研究进展

夏季热应激给奶牛的产奶性能、瘤胃健康、免疫机能和繁殖性能等方面带来了一定的负面影响,严重危害着奶牛健康,成为制约我国奶业健康发展的重要因素之一。本文探讨了热应激发生的原因及对奶牛生产的负面影响,并从品种选育、饲养管理和营养调控等方面总结了减缓热应激相关措施的研究进展,以期为奶牛热应激的综合防控提供参考。

基于视频和BCE-YOLO模型的奶牛采食行为检测

【目的】动物采食行为是一个重要的动物福利评价指标,本研究旨在解决复杂饲养环境下奶牛采食行为识别精度差、特征提取不充分的问题,实现对奶牛采食行为的自动监控。【方法】本文提出了一种基于改进BCEYOLO模型的识别方法,该方法通过添加BiFormer、CoT、EMA 3个增强模块,提高YOLOv8模型特征提取的能力,进一步与优于Staple、SiameseRPN算法的Deep SORT算法结合,实现对奶牛采食时头部轨迹的跟踪。在奶牛采食过程的俯视和正视视频中提取11288张图像,按照6∶1比例分为训练集和测试集,构建采食数据集。【结果】改进的BCE-YOLO模型在前方和上方拍摄的数据集上精确度分别为77.73%、76.32%,召回率分别为82.57%、86.33%,平均精确度均值分别为83.70%、76.81%;相较于YOLOv8模型,整体性能提升6~8个百分点。Deep SORT算法相比于Staple、SiameseRPN算法,综合性能提高1~4个百分点;并且改进的BCE-YOLO模型与Deep SORT目标跟踪算法结合良好,能对奶牛采食行为进行准确跟踪且有效地抑制了奶牛ID(Identity document)的变更。【结论】本文提出的方法能有效解决复杂饲养环境下奶牛采食行为识别精度差、特征提取不充分的问题,为智能畜牧与精确养殖提供重要参考。

高产奶牛饲养管理技术措施探讨

保证高产奶牛的良好发育和理想产奶量,需注重各阶段的饲养管理。高产奶牛饲养管理有效措施包括选择优秀牛种,以改良本奶牛场的牛群结构;在奶牛不同生长阶段,科学合理选择不同的饲养方式;产奶牛需保持足够的运动量、固定的饲喂模式、预防热应激、清洁的挤奶卫生和牛舍卫生、慎重淘汰高产奶牛等。本文主要论述高产奶牛饲养管理中的技术要点,旨在为养殖户有序开展高产奶牛饲养活动,提升整体饲养水平提供参考。

叶酸和甜菜碱对围产期奶牛泌乳性能、乳脂肪酸组成、养分消化和瘤胃发酵的影响

本试验旨在研究叶酸(FA)和甜菜碱(BT)对围产期奶牛泌乳性能、乳脂肪酸组成、养分消化和瘤胃发酵的影响。选取预产日期、胎次、体重和上一泌乳期平均产奶量相近的围产期荷斯坦奶牛40头,随机分为4组,每组10头牛。对照组饲喂基础饲粮,FA组、BT组和FA+BT组分别在基础饲粮中添加135 mg/d的FA、100 g/d的BT和135 mg/d的FA+100 g/d的BT。预试期10 d,正试期42 d。结果表明:1)FA组和FA+BT组的实际乳、4%乳脂校正乳、乳脂肪、乳蛋白质产量以及乳脂率和饲料效率(FE)显著高于对照组(P<0.05)。BT组的4%乳脂校正乳、乳脂肪、乳蛋白质产量以及乳脂率和FE显著高于对照组(P<0.05)。2)FA组乳脂肪中葵酸(C10∶0)、月桂酸(C12∶0)、肉豆蔻酸(C14∶0)、棕榈一稀酸(C16∶1)、亚麻酸(C18∶3)和总4~15碳脂肪酸含量显著高于对照组(P<0.05),硬脂酸(C18∶0)含量显著低于对照组(P<0.05)。BT组和FA+BT组乳脂肪中C10∶0、C12∶0、C14∶0、C16∶1、亚油酸(C18∶2)、C18∶3和总4~15碳脂肪酸含量显著高于对照组(P<0.05),C18∶0和总>16碳脂肪酸含量显著低于对照组(P<0.05)。3)FA组、BT组和FA+BT组的干物质(DM)、有机物(OM)、中性洗涤纤维(NDF)和酸性洗涤纤维(ADF)表观消化率显著高于对照组(P<0.05)。4)FA组、BT组和FA+BT组的瘤胃总挥发性脂肪酸(TVFA)、乙酸和丁酸含量及乙酸/丙酸显著高于对照组(P<0.05)。由此可见,围产期奶牛饲粮中添加FA和BT能改善泌乳性能和养分消化,而且FA和BT共同添加对改善泌乳性能比单独添加更有效。

泌乳前期奶牛在不同采食水平下产奶性能的差异比较

[目的]探究泌乳前期奶牛在不同采食水平下产奶性能的差异。[方法]本试验挑选45头处于泌乳前期、体况和胎次相近的健康荷斯坦牛,按照采食量低采食量组、中采食量组和高采食量组,预试验7 d,正式试验28 d。[结果]与低采食量组相比,中采食量组和高采食量组奶牛的平均日采食量显著升高(P<0.05);高采食量组奶牛的采食速度、产奶量和4%标准乳产量显著升高(P<0.05),乳中尿素氮含量显著降低(P<0.05)。中采食量组和高采食量组奶牛瘤胃内pH值、氨态氮、乙酸、丙酸及总挥发性脂肪酸含量显著升高(P<0.05)。中采食量组和高采食量组奶牛血清中白蛋白、总蛋白含量显著升高(P<0.05);高采食量组奶牛血清中尿素氮含量显著升高(P<0.05)。[结论]泌乳前期奶牛在不同采食水平下产奶性能不同,采食量越高,产奶性能越好,瘤胃发酵功能越强,机体体况越佳。

Effect of Fruit-flavored Agent and Green Pigment on Production Performance of Simmental Cows

[Objective] The paper was to increase utilization rate of roughage to solve current shortage of forage and improve production performance of Simmental cow.[Method] Fruit-flavored agent and green pigment were added in roughage of Simmental cow respectively,and their effects on feed intake,milk production and milk quality of Simmental cow were studied.[Result] After adding fruit-flavored agent and green pigment,the feed intakes of Simmental cow were increased by 30.69% and 12.27%,while the milk productions were increased by 1.74 and 2.25 kg,respectively,and the differences were all significant（P 〈 0.05）.[Conclusion]Adding 0.3% and 0.1% fruit-flavored agent and green pigment in roughage of Simmental cow could significantly improve feed intake and milk production,which can also delay the decline of milk production during late lactation stage and improve milk quality.

奶牛胃肠道甲烷减排措施研究概述

在牛奶生产的过程中,奶牛胃肠道是甲烷产生的主要场所。本文概述了日粮NFC、淀粉、蛋白、纤维、脂肪、添加剂等营养调控措施和DMI、健康、牛群结构等管理措施及产奶量、饲料转化效率和反刍行为等遗传选择措施对奶牛胃肠道甲烷减排的影响。

益生菌发酵饲料对奶牛肠道菌群和生产性能的影响

[目的]研究益生菌发酵饲料对奶牛肠道菌群和生产性能相关指标的影响。[方法]采用单因子随机试验设计,将200头年龄、胎次和平均产奶量相近的荷斯坦牛均分成4组:对照组(基础日粮)、低剂量组(基础日粮+1%的益生菌)、中剂量组(基础日粮+2%的益生菌)、高剂量组(基础日粮+4%的益生菌)。每组设置5个重复,每个重复奶牛数量相同。预试验期7 d,正式试验期35 d。正式试验期间统计分析奶牛的平均日增重、平均日产奶量、料奶比,检测第35d粪便的大肠杆菌、沙门氏菌、乳酸菌数量。[结果]与对照组相比,低剂量组、中剂量组、高剂量组平均日采食量、平均日产奶量显著提高;低剂量组、中剂量组、高剂量组料奶比显著降低;低剂量组、中剂量组、高剂量组粪便的大肠杆菌、沙门氏菌数量显著降低;低剂量组、中剂量组、高剂量组粪便的乳酸菌数量显著升高。此外,高剂量组平均日采食量、平均日产奶量、粪便的乳酸菌数量最高,料奶比及粪便的大肠杆菌和沙门氏菌数量最低。[结论]益生菌发酵饲料能有效提高奶牛生长性能,改善肠道菌群,最佳使用量4%。

中药组方配合西医对奶牛真胃炎治疗效果、排泄和采食的影响

[目的]研究中药组方配合西医对奶牛真胃炎治疗效果、排泄和采食的影响。[方法]以山东滨州市某规模化奶牛养殖场随机选择166头患真胃炎的成年泌乳奶牛为研究对象,年龄、胎次、体态等相近。按照随机分组方法,分为观察组和对照组。对照组采用西医疗法,观察组在此基础上配合中药组方治疗。治疗7 d,比较两组治疗效果和治疗后排泄、采食情况。[结果]观察组奶牛治疗效果明显高于对照组,总治愈率高于对照组(P<0.05);观察组奶牛治疗后排泄次数、排泄物性状均恢复正常,正常率高于对照组(P<0.05);观察组奶牛采食量高于对照组(P<0.05),观察组奶牛反刍次数均在正常范围内,优于对照组(P<0.05)。[结论]对真胃炎奶牛采用中药组方配合西医治疗,可提高治疗效果,促进尽快恢复正常排泄和采食,有推广优势和价值。

基于Swin-Unet的奶牛饲料消耗状态监测方法

【目的】针对监控图像中饲料区域结构较长、边界模糊,且形状与尺寸复杂多变等特点,本研究旨在更准确地分割饲料残余区域与消耗区域,以达到准确监测饲料消耗状态的目的。【方法】本研究提出了基于Swin-Unet的语义分割模型,其在Swin Transformer块的开始阶段应用ConvNeXt块,增强模型对特征信息的编码能力,以提供更好的特征表示,并利用深度卷积替换线性注意力映射,以提供局部空间上下文信息。同时提出了新颖的宽范围感受野模块来代替多层感知机,以丰富多尺度空间上下文信息。此外在编码器的开始阶段,将线性嵌入层替换为卷积嵌入层,通过分阶段压缩特征,在块之间和内部引入更多的空间上下文信息。最后引入多尺度输入策略、深度监督策略,并提出了特征融合模块,以加强特征融合。【结果】所提出方法的平均交并比、准确率、F1分数与运行速度分别为86.46%、98.60%、92.29%和23帧/s,相较于Swin-Unet,分别提高4.36、2.90、0.65个百分点和15%。【结论】基于图像语义分割的方法应用于饲料消耗状态的自动监测是可行的,该方法通过将卷积引入Swin-Unet,有效地提高了分割精度与计算效率,对提升生产管理效率具有重要意义。

基于机器学习的奶牛饲料消耗状态预测模型

饲料作为奶牛重要的营养来源,预测饲料消耗状态对于保障奶牛的健康和提高生产管理效率具有重要意义。然而,由于饲料消耗状态数据呈现出非线性和非平稳的特点,导致预测精度较低。为解决此问题,该研究基于经验模态分解(empirical mode decomposition,EMD)和长短期记忆网络(long short-term memory,LSTM),提出了组合改进的自适应噪声完全集成经验模态分解(improved complete ensemble empirical mode decomposition with adaptive noise,ICEEMDAN)、随机森林(random forest,RF)与改进的LSTM(improved LSTM,ILSTM)的模型,即ICEEMDANRF-ILSTM,来预测饲料消耗状态。其通过调整遗忘门的输出值范围以增强模型的特征学习能力。首先,使用ICEEMDAN对饲料消耗状态数据进行分解,得到多个相对平稳的分量。其次,考虑到每个分量具有不同的特性,采用不同的方法来建模不同的分量,以进一步提升预测效果。具体而言,为了提升模型的精度以及泛化能力,使用RF建模频率最高的分量;同时,使用ILSTM建模其余分量,以捕获序列数据中的长期依赖性。最后,将所有分量的预测结果相加得到最终的预测结果。基于自建数据集的试验结果表明,ICEEMDAN-RF-ILSTM对于饲料消耗状态预测具有较高的准确度,其决定系数R2、平均绝对百分比误差与均方根误差分别为0.993、2.576%和0.596%,表明其能有效预测饲料消耗状态,同时其性能优于ICEEMDAN-LSTM模型。该研究为评估饲料消耗状态提供了可行的方法,可为制定调度决策提供了科学的技术支持,并为牧业智能化建设提供借鉴。